Linking Digital Twin Design and Ontologies with Model-Driven

Engineering: Application to Railway Infrastructure

Alexis Chartrain

1,2

, Gilles Dessagne

, No

el Haddad

and David R. C. Hill

SNCF R

ESEAU, Direction G

erale Industrielle & Ing

enierie, F-93210 Saint-Denis, France

Universit

e Clermont – Auvergne, CNRS, Clermont – Auvergne INP, Mines de Saint-

Etienne,

LIMOS UMR CNRS 6158, F-63000 Clermont-Ferrand, France

{alexis.chartrain, gilles.dessagne, noel.haddad}@reseau.sncf.fr, david.hill@uca.fr

Keywords:

Digital Twin, Information System, Metamodels, Meta-Object Facility (MOF), Model-Driven Architecture

(MDA), Model-Driven Engineering (MDE), Object-Oriented Approach, Ontologies, Railway System.

Abstract:

In this position paper, we argue that ontologies, combined with Model-Driven Engineering (MDE) approach

and the object-oriented approach, could be leveraged to produce model-driven Digital Twins. This paper

presents our framework for the production of model-driven Digital Twins using the aforementioned approach.

Despite the interest in existing frameworks, this paper offers a new perspective that could help pave the way

towards a future standardized, generic framework and shows insights of application at the French Railway

Infrastructure Manager, SNCF R

ESEAU.

1 INTRODUCTION

The concept of “Digital Twin” ﬁrst emerged in the

early 2000’s and has seen signiﬁcant growth since

2015, as evidence by the increasing number of an-

nual publications on the topic, particularly from the

period between 2015 and 2017 (Barricelli et al., 2019;

Semeraro et al., 2021). Indeed, Digital Twins (DTs)

are seen as promising tools for monitoring, simulat-

ing, optimizing, and predicting the behaviour of a

real-world system or product, remotely from a virtual

counterpart that mirrors the actual system or prod-

uct. These capabilities are highly valued in cutting-

edge sectors, which is why DTs are gaining traction in

manufacturing (e.g., Industry 4.0, aerospace), health-

care (e.g., hospitals), transportation (shipping, avia-

tion, maritime, railways) (Barricelli et al., 2019; Se-

meraro et al., 2021; De Donato et al., 2023).

However, deﬁning, designing, and developing a

DT is by no means straightforward. One major chal-

lenge is to deal with the numerous deﬁnitions stated

for the DT concept, which vary across sectors, appli-

cations, and use-cases. As a result, there is little con-

sensus on a universal deﬁnition of this concept within

the community, both in industry and academia (Tao

and Qi, 2019; Adamenko et al., 2020; Zhang et al.,

2021; VanDerHorn and Mahadevan, 2021; De Donato

et al., 2023; Chartrain et al., 2024). Nevertheless,

a clear deﬁnition is crucial to determine what needs

to be designed and developed. Furthermore, creat-

ing a DT requires models (Tao and Qi, 2019; Char-

train et al., 2024); however despite the existence of a

few frameworks, standards and methods for design-

ing and producing DTs (Segovia and Garcia-Alfaro,

2022; Haße et al., 2022), even fewer of these ap-

proaches are model-driven (Zhang et al., 2021).

In computer science, ontologies are considered

as speciﬁc, often formalized, conceptualizations ac-

counting for a particular view or vision of the world

(e.g., reality, domain of discourse) which are shared

among a group of people (Gruber, 1995; Guarino,

1998; Maedche, 2002; Gruber, 2008). They are pre-

sented as structured corpus of concepts in relation-

ship (i.e., semantic or taxonomic relations), modeled

in a language allowing further seamless exploitation

by computers. They are often represented by knowl-

edge graphs and formalized through languages such

as the Web Ontology Language (OWL), the Resource

Description Framework (RDF), or the Uniﬁed Mod-

eling Language (UML). A computer ontology is al-

ways built for a domain, but less frequently for a

set of domains of knowledge. Ontologies offer high-

potential solutions for analyzing domains, systems,

and products by formalizing their inherent concepts

and interrelationships, thereby providing the founda-

tion for related unifying models. DTs of these sys-

tems or products, considered as their digital repre-

sentations (i.e., a speciﬁc type of model) conform to

Chartrain, A., Dessagne, G., Haddad, N. and Hill, D.

Linking Digital Twin Design and Ontologies with Model-Driven Engineering: Application to Railway Infrastructure.

DOI: 10.5220/0013090200003838

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 16th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2024) - Volume 2: KEOD, pages 265-276

ISBN: 978-989-758-716-0; ISSN: 2184-3228

265

their related meta-models, following B

ezivin’s works

on Model-Driven Engineering (B

ezivin and Gerb

2001; B

ezivin, 2004; B

ezivin, 2005).

In this paper, we argue that ontologies com-

bined with Model-Driven Engineering (MDE) stan-

dards and techniques, and the object-oriented ap-

proach could be leveraged to produce model-driven

DTs. This paper aims at presenting our proposal of

framework for the production of model-driven DTs

with the previously mentioned approach. Our goal is

not to dismiss existing frameworks, but to offer a new

perspective that could help pave the way toward a fu-

ture standardized, generic framework.

In section 2, we present an overview of DTs and

ontologies as part of the state of the art. We also pre-

cise the questions addressed in this paper regarding

the production of DTs. Then, in section 3, we intro-

duce our framework which aimed at facilitating the

production of DTs. This framework includes deﬁni-

tions of the key concepts, basic object-oriented prin-

ciples, Model-Driven Engineering (MDE) standards,

ontologies, and the strong links between these ele-

ments. We also discuss the application of this frame-

work at SNCF R

ESEAU, the French Railway Infras-

tructure Manager, within the context of producing a

railway system DT as part of the company Informa-

tion System. We give a very concrete example of ap-

plication in railways dealing with turnouts and associ-

ated measurements, especially from the perspective of

monitoring point machines current consumption with

our DT. Finally, we conclude this paper in section 4.

2 DIGITAL TWIN AND

ONTOLOGIES: AN OVERVIEW

In this section, we present a brief state of the art re-

garding the two main topics of our paper: on the one

hand the concept of DT and, on the other hand on-

tologies in computer science. For both of these top-

ics, we proceed to a short overview of their historical

background and a review of existing deﬁnitions in the

literature including their main characteristics.

2.1 Where Does the Concept of Digital

Twin Comes From?

2.1.1 Historical Background and Deﬁnitions

The scientiﬁc community acknowledges Michael

Grieves as one of the ﬁrst – if not the ﬁrst – author that

introduced the concept of “Digital Twin” in the liter-

ature (Kritzinger et al., 2018; Barricelli et al., 2019;

Madni et al., 2019; Semeraro et al., 2021; Segovia and

Garcia-Alfaro, 2022). Roots of this concept were in-

troduced in 2002 by Grieves, in his presentation slide

show for the formation of a Product Lifecycle Man-

agement (PLM) center, and more precisely in the well

known slide “Conceptual Ideal for PLM” (Grieves

and Vickers, 2017; Barricelli et al., 2019). Although

the concept was not yet designated as “Digital Twin”

at that time, all the elements of Grieves’ deﬁnition

were already stated (i.e., a real space, a virtual space,

and a data ﬂow that connects the two spaces together)

(Grieves and Vickers, 2017). These elements, referred

as “Mirrored Spaces Model”, were then presented in

Grieves’ courses on PLM at the University of Michi-

gan in 2003. Grieves himself attributes to John Vick-

ers the origin of the term “Digital Twin”. Accord-

ing to the former, the latter has ﬁrst proposed it in

the framework of their joint research works. The au-

thors have reﬁned their concepts over the years from

“Conceptual Ideal for PLM” to “Digital Twin”, by

the way of “Mirrored Spaces Model” and “Informa-

tion Mirroring Model”. However, the essence of these

concepts remained mostly the same with the follow-

ing three main aspects: (1) a real space containing

physical products, (2) a virtual space containing vir-

tual products, and (3) a connection between the two

spaces, tying them together through an information

exchange achieved by a data ﬂow (Grieves, 2015;

Grieves and Vickers, 2017).

In 2012, ten years after Grieves’ preliminary

works on the DT concept, NASA proposed their def-

inition in regards of their own needs: “A Digital Twin

is an integrated multiphysics, multiscale, probabilis-

tic simulation of an as-built vehicle or system that

uses the best available physical models, sensor up-

dates, ﬂeet history, etc., to mirror the life of its corre-

sponding ﬂying twin.” (Glaessgen and Stargel, 2012;

Barricelli et al., 2019).

In 2018, Kritzinger et al. reﬁned the deﬁnition

of the DT concept with a proposal for a categoriza-

tion of the different integration levels of a physical

system with a corresponding virtual system (i.e., the

twin). According to the latter, three levels could be

distinguished, respectively named digital model, dig-

ital shadow, and digital twin (Kritzinger et al., 2018).

We detail each one of them thereafter.

A digital model is a digital representation of an ex-

isting or planned physical system without any au-

tomated information ﬂow between the physical

system and its corresponding representation. In

this case, digital data regarding the physical sys-

tem could only be set manually in the virtual sys-

tem; consequently a change of state in one of the

system has no effect on the other.

DTO 2024 - Special Session on Ontologies for Digital Twin

266

A digital shadow is a virtual system (or digital

model) being automatically updated with a data

ﬂow containing information about the physical

system (e.g., gathered with sensors) allowing a

possible control – in the monitoring sense – of the

physical system through the corresponding virtual

system.

A digital twin is a virtual system being both capable

of commanding (i.e., transmission of instructions)

and controlling (i.e., monitoring of up-to-date sta-

tuses such as in the digital shadow case) a physical

system.

To sum up Kritzinger’s position: a digital model is

a digital representation disconnected from any phys-

ical system; a digital shadow imply a one-way infor-

mation ﬂow from a physical system to a correspond-

ing digital representation. Finally only the digital

twin runs a double-way information ﬂow between a

physical system and a corresponding digital represen-

tation (Kritzinger et al., 2018; Segovia and Garcia-

Alfaro, 2022). In addition, (Aheleroff et al., 2021)

based their reasoning on the categorization proposed

in (Kritzinger et al., 2018), but add one more kind:

the digital twin predictive that uses the digital repre-

sentation to process data and perform simulations or

machine learning applications to make predictions.

Following from this position, (Wright and David-

son, 2020), defended in 2020 a clear distinction be-

tween the two notions of “model” and “Digital Twin”.

They argued that the existence of a physical system is

an indispensable condition for a related digital repre-

sentation to be designated by the term “Digital Twin”.

Otherwise, without any physical counterpart, the digi-

tal representation is simply designated as a “plain dig-

ital model”. However, this point of view should be

put in perspective with the PLM vision proposed by

(Grieves and Vickers, 2017) with different types of

DTs (e.g., Digital Twin Prototype, Digital Twin In-

stance) and the approach with further propose, both

embracing the whole lifecycle of the system within

the DT.

2.1.2 Characteristics and Functionalities

According to our previous analysis stated in (Char-

train et al., 2024), we propose to arrange the most

common characteristics and functionalities of DTs

described in the existing literature within the three fol-

lowing aspects:

1. A digital representation of a precise, designated,

system or product – often assumed as physical –

that interest a group of stakeholders, aiming at

digitally represent this system/product, possibly

in an up-to-date or pseudo real-time manner (Bar-

ricelli et al., 2019; Madni et al., 2019).

2. An information ﬂow between an actual sys-

tem/product and a corresponding digital represen-

tation aiming at controlling (i.e., monitoring) and

eventually commanding the actual system from

its associated virtual counterpart (Grieves, 2015;

Kritzinger et al., 2018; Tao and Qi, 2019; Segovia

and Garcia-Alfaro, 2022).

3. Simulations, predictions, and decisions achieved

based on the current available knowledge of the

actual system (e.g., digital models, data) aim-

ing at anticipate the behaviour of this actual sys-

tem/product by predicting its future likely sta-

tuses as effectively as possible. This is per-

formed thanks to statistics, simulations, optimiza-

tions, and even artiﬁcial intelligence tools, based

on the digital representation of the actual sys-

tem/product (Barricelli et al., 2019; Segovia and

Garcia-Alfaro, 2022).

In this paper, we argue that the integration of the ac-

tual system within the corresponding DT and the pre-

dictions achieved thanks to the latter about the former,

both rely on the digital representation containing all

the necessary information to describe, manage, mon-

itor, control, and predict the behaviour of this actual

system of interest throughout its whole lifecycle. As a

result, the digital representation is the cornerstone of

the DT and will constitute our main focus from now

on.

2.2 Ontology or Ontologies?

2.2.1 Historical Background and Deﬁnitions

The term “Ontology” is a combination of the Ancient

Greek words ontos and logos, the former meaning

“being” or “what is”, and the latter meaning “treatise”

or “discourse”. Ontology is a branch of philosophy

that originates from Aristotle’s metaphysics, a disci-

pline interested in the study of the being considering

its intrinsic nature, characteristics, and organization;

broadly, the study of existence and reality focusing

on “what exists” (Maedche, 2002; Gruber, 2008).

We consider that the ﬁeld of Ontology allows

practitioners of the discipline (e.g., philosophers) to

produce ontologies. Guarino, then Maedche high-

lighted the paramount importance of differentiating

the Ontology (i.e., the philosophical discipline) from

an ontology – or ontologies in its plural form – that

designate “a particular system of categories account-

ing for a certain vision of the world” (Guarino, 1998;

Maedche, 2002). As for Gruber, he thinks an ontol-

Linking Digital Twin Design and Ontologies with Model-Driven Engineering: Application to Railway Infrastructure

267

ogy in philosophy in a similar manner, that is as “a

systematic account of Existence” (Gruber, 1995).

Furthermore, nowadays ontologies are not only

contained to the ﬁeld of philosophy anymore; it has

indeed considerably expanded to the ﬁeld of com-

puter science over the past three to four decades.

We have recently seen various applications in Artiﬁ-

cial Intelligence (e.g., machine learning, deep learn-

ing, Large Language Models), Computational Lin-

guistics, Knowledge Engineering (e.g., knowledge

representation, data/information modeling, object-

oriented analysis, shared digital libraries of reusable

concepts/knowledge components, Information Sys-

tems) and Database Theory (e.g., heterogeneous data

management, multi-database systems interoperabil-

ity) (Guarino, 1998; Gruber, 1995; Gruber, 2008).

Like philosophers, computer scientists are also pro-

ducing ontologies to organize and formalize their

knowledge, for example to better understand a do-

main. Computer scientists ﬁrst and foremost use them

to make high-level languages computable. In addi-

tion, Knowledge Engineering and Database Theory

are promising ﬁelds that should be considered for the

design and the production of DTs, especially in the

framework of Information Systems. In the light of the

above, the need of better approaching the concept of

an ontology with a focus on computer science clearly

appears; therefore we establish a corpus of deﬁnitions

of an ontology in computer science in the next section.

2.2.2 The Concept of Ontology in Computer

Science

Probably one of the most famous deﬁnitions of an

ontology in computer science, is the following one

stated by Gruber: “an ontology is an explicit speci-

ﬁcation of a conceptualization” (Gruber, 1995; Gru-

ber, 2008). The so-called “conceptualization” is “an

abstract, simpliﬁed view of the world that we wish to

represent for some purpose” or “an abstract, simpli-

ﬁed view of a domain of discourse” (Gruber, 1995;

Gruber, 2008; Aßmann et al., 2006; Valiente et al.,

2011). According to Guarino, then Maedche: “an on-

tology refers to an engineering artifact, constituted by

a speciﬁc vocabulary used to describe a certain real-

ity, plus a set of explicit assumptions regarding the in-

tended meaning of the vocabulary words” (Guarino,

1998; Maedche, 2002). As for Ben Sta et al., they

deﬁne an ontology as “an explicit and formal shared

abstract view of a part of the real world. This view

is described by a whole [set] of tools as a vocabulary

formed of concepts, relations, axioms and rules of in-

ference” (Ben Sta et al., 2005).

The nature and the characteristics of an ontology

in computer science are for the most part, acknowl-

edged in the literature: it is a conceptualization with

shared, explicit, and possibly formal aspects. The dif-

ferent kinds of components of ontologies are: con-

cepts, relations, and axioms (Gruber, 1995; Ben Sta

et al., 2005; Aßmann et al., 2006; Gruber, 2008; Va-

liente et al., 2011).

Furthermore, ontologies are classiﬁed in several

levels depending on their scopes. According to (Guar-

ino, 1998) then (Maedche, 2002), there are: (1) top-

level ontologies, also known as upper-level ontolo-

gies, which describe common high-level concepts of

a general vocabulary (e.g., space, time, matter, object,

event, action) independent from any domain, task, or

problem; (2) domain ontologies and tasks ontologies,

which describe concepts related to the vocabulary of

a speciﬁc domain (e.g., railways) or a given task, by

specializing concepts from a top-level ontology; and

(3) application ontologies, linking domain and task

ontologies by describing the roles played by domain

entities through the performance of a given process,

activity, or task.

2.3 Still Problems

As stated in the introduction of this paper and already

highlighted in one of our previous publication, there

are still remaining open questions about DTs (Char-

train et al., 2024). First, it appears that hardly any

clear consensus on a generic, universal, deﬁnition of

the concept of Digital Twin exist to date among re-

searchers and practitioners in the community (Tao and

Qi, 2019; Barricelli et al., 2019; Adamenko et al.,

2020; Zhang et al., 2021; VanDerHorn and Mahade-

van, 2021; De Donato et al., 2023). To illustrate this

problem, let’s take the following example: among the

75 papers reviewed in (Barricelli et al., 2019), only 31

provided a deﬁnition of the DT concept, counting 29

different deﬁnitions in total. Depending on the deﬁ-

nitions provided in the papers, these papers were then

categorized in the 6 following topics: (1) “Integrated

system” (2) “Clone, counterpart” (3) “Ties, links”

(4) “Description, construct, information” (5) “Simu-

lation, test, prediction” (6) “Virtual, mirror, replica”

(Barricelli et al., 2019). In addition, based on a cor-

pus of 150 papers and an analysis of 30 deﬁnitions of

the DT concept, (Semeraro et al., 2021) proposed a

classiﬁcation of these deﬁnitions following ﬁve clus-

ters: (1) “Lifecycle”, (2) “Cyber-Physical Systems”,

(3) “Real and virtual spaces in loop”, (4) “Behaviour

modeling of a physical space”, (5) “Virtual system”

i.e., replication of a physical system. In (Chartrain

et al., 2024), we conjectured that numerous deﬁni-

tions of the DT concept might exist because it is fore-

most approached with use-cases (e.g., lifecycle man-

DTO 2024 - Special Session on Ontologies for Digital Twin

268

agement, control/command, predictive maintenance)

and not by deﬁning its very nature; according to Bar-

ricelli et al.: “literature works have never described

in detail the characteristics of a generic DT. Indeed,

each state-of-the-art paper concentrates on the devel-

opment of few components of DTs” (Barricelli et al.,

2019). In the next section, we will propose a generic

deﬁnition that we hope broad enough to embrace as

many cases as possible, and as reusable as possible.

We will then reﬁne this deﬁnition to suit our objective

regarding the production of DTs with our framework.

Secondly, it appears as well that any DT requires

models to be produced (Tao and Qi, 2019; Tao et al.,

2022). In (Chartrain et al., 2024), we quoted Tao

and Qi: “To build a digital twin of an object or sys-

tem, researchers must model its parts” (Tao and Qi,

2019). We also underlined the importance of concep-

tual models (i.e., ontologies) since they contain all the

necessary semantics (i.e., concepts and relationships)

to create digital representations of systems accurately

from the desired point of view in DTs. Pieces of

information within digital representations are stored

through data, according to the structure of concepts

in related conceptual models.

A few frameworks and methods currently exist to

produce DTs: we could for instance cite the DT Ar-

chitecture Reference Model (Aheleroff et al., 2021),

the design principles for shared DTs in distributed

systems (Haße et al., 2022), the proposal for design,

modeling, and implementation (Segovia and Garcia-

Alfaro, 2022). Also, ISO standards were recently

published: ISO 23247:2021 focused on automation

& manufacturing, then ISO/IEC 30173:2023 with a

larger scope. However, these frameworks, methods,

and norms are only on early stages of maturation;

therefore there is hardly any consensus among the

community about which ones should be used in each

case in order to design and produce a DT (Tao and

Qi, 2019; Zhang et al., 2021; De Donato et al., 2023;

Segovia and Garcia-Alfaro, 2022). Regarding model-

driven DTs, only Zhang et al. describes a model-

based approach using MDE techniques to design a DT

considering its whole lifecycle (Zhang et al., 2021).

However, the latter doesn’t focus much on model-

ing choices i.e., what should be represented and how.

Based on requirements, the production of models con-

stitutes just a step in the whole framework proposed

to design a complete DT in (Zhang et al., 2021). Our

framework presented in the next section is our contri-

bution to mitigate this lack and initiate a bridge be-

tween DT designing and ontology engineering.

3 LINKING DIGITAL TWIN

DESIGN AND ONTOLOGIES

WITH MODEL-DRIVEN

ENGINEERING

In this section, we present our proposal of metamod-

eling framework regarding the design and the produc-

tion of a Digital Twin as part of an Information Sys-

tem, using ontologies, the object-oriented approach,

and existing standards coming from Model-Driven

Engineering (MDE). We ﬁrst introduce the object-

oriented approach and the standards we are using,

secondly we propose our deﬁnitions of the key con-

cepts given this context and, thirdly we present our

complete proposal. Finally, we provide an example

of application at SNCF R

ESEAU, the French Railway

Infrastructure Manager.

3.1 World Modeling Approaches

3.1.1 Object-Oriented Approach

The object-oriented approach relies on the concepts

of instances and classes (Dahl and Nygaard, 1966).

The instance is constituted of a ﬁnite set of attributes

and a ﬁnite set of methods; each attribute is deﬁned

by its name (i.e., a property name) and its associated

value; similarly, each method is deﬁned by its name

(i.e., a function name) and its returned value.

In Object-Oriented Programming (OOP), an in-

stance is generated from a class that is, a common

structure containing generic attributes and methods

for instances. For that matter, a class could be consid-

ered as a model for instances: “A class forms a model

for the creation of instances which are only individual

representations of this model.” (Hill, 1996). Also, we

could draw a strong link between a class and a concept

in an ontology, considering their close deﬁnitions and

features. The concepts and relationships expressed in

computer science ontologies matches perfectly what

we need to represent through models, and more par-

ticularly when going to an upper modeling level, fur-

ther introduced with the concept of metamodel.

In the DT engineering perspective of our pro-

posal, our goal is to instantiate the digital represen-

tation from classes and to design the classes in ac-

cordance with the concepts inherent to the actual sys-

tem/product with an ontological approach. To do so,

we use existing Model-Driven Engineering (MDE)

standards provided by the Object Management Group

(OMG), further discussed thereafter.

Linking Digital Twin Design and Ontologies with Model-Driven Engineering: Application to Railway Infrastructure

269

3.1.2 Object Management Group Model-Driven

Standards

Our framework requires the use of two OMG stan-

dards: the Meta-Object Facility (MOF) on the top of

the metamodeling architecture and the Model-Driven

Architecture (MDA). It also requires the famous four-

layer metamodeling architecture, commonly illus-

trated with the “meta-pyramid”.

This metamodeling architecture contains four lev-

els respectively called M0, M1, M2, and M3 (B

ezivin

and Gerb

e, 2001). M0 is the ground level in which no

models are involved (i.e., 0 model = M0), it simply

designates the “real world” in which real systems or

products are contained. M1 is the ﬁrst level of mod-

eling, embracing models of the real world. M2 is the

second level of modeling, concerned with metamod-

els as models of languages to produce models. Fi-

nally, M3 is the last level of modeling which contains

the unique metametamodel able to describe itself. It

is the MOF within the OMG standard (B

ezivin and

Gerb

e, 2001; B

ezivin, 2005). The M3 layer allows all

metamodels in M2 to be compatible with each other.

In our framework, we stick to B

ezivin’s 3+1 revis-

ited organization in which models in M1 represent a

real system in M0, models in M1 conform to meta-

models in M2, and lastly metamodels in M2 conform

to the MOF in M3, the MOF being self-conformant

ezivin, 2005). This reﬂexivity is known as metacir-

cularity by specialists.

The MDA is a standard enabling the object-

oriented analysis, design, and programming in soft-

ware development, allowing an uncoupling of the re-

sults of these activities in three different kinds of mod-

els: (1) Computation Independent Models (CIM) that

are requirement models presenting an analysis of a

domain or a real-world system, through the formal-

ization of related concepts which could be drawn with

an ontology; (2) Platform Independent Models (PIM)

stating software design choices and solutions based

on the analysis in the CIM; and (3) Platform Spe-

ciﬁc Models (PSM) constituting object-oriented im-

plementation of the former models. We position all

these models at level M1 in the metamodeling archi-

tecture, this view is shared (Aßmann et al., 2006; Va-

liente et al., 2011). During the whole development

process, we consider that a CIM is transformed in one

or several PIMs, and that each PIM is transformed in

one or several PSMs, depending on the project needs.

This transformation relationship between models is

well documented in Favre’s works on MDE (Favre

and Nguyen, 2005).

3.2 Our Deﬁnitions of the Key Concepts

3.2.1 Digital Twin

To deﬁne what we have to design and produce, we

ﬁrst start by deﬁning the DT concept in a broad sense

that is, as an up-to-date digital representation of a sys-

tem of interest. This representation is shared among a

group of people (e.g., stakeholders working in collab-

oration). Furthermore, it can embrace all the neces-

sary pieces of information for describing (e.g., func-

tional models), monitoring/controlling, (e.g., statuses,

measures, alerts, warnings), commanding (e.g., or-

ders, instructions), and simulating (e.g., simulation

models, data) the actual system.

We consider that the integration of the physical

system within the corresponding digital representa-

tion depend on our use-cases and therefore may even

change over time. For example, the DT could contain

information about a system not yet produced, conse-

quently without an immediate physical reality e.g.,

Grieves’ Digital Twin Prototype (Grieves and Vick-

ers, 2017). We give another example: a DT could ﬁrst

be designed to monitor a system, then upgraded later

to perform control/command, and ﬁnally augmented

afterwards to perform simulations based on the dig-

ital representation to inform decisions related to the

actual real-world system. The digital representation

should be capable of addressing all use-cases related

to every phase of the actual system lifecycle during

which we aim to work with the DT.

Hence, we argue that the heart of a DT lies in the

digital representation in the ﬁrst place. As a result, we

propose to deﬁne a DT as follows:

Deﬁnition 1. Generic deﬁnition. A Digital Twin (DT)

is a shared up-to-date digital representation of a sys-

tem of interest (Chartrain et al., 2024).

Considering now an implementation perspective,

and using the object-oriented approach we have pre-

viously introduced:

Deﬁnition 2. Object-oriented technical deﬁnition. A

Digital Twin (i.e., as deﬁned in 1) could be imple-

mented with a set of instances generated from a global

and systemic class model. This representation could

be stored in repositories and be accessible through

services (Chartrain et al., 2024).

Of course, this second deﬁnition is less generic

and may certainly not be reusable in every case;

it assumes that the design of the considered DT is

achieved with the object-oriented approach. When

using a model-driven approach, choices of other tech-

nical spaces are possible.

DTO 2024 - Special Session on Ontologies for Digital Twin

270

3.2.2 Concepts of (Meta-)Model, Megamodel,

and Ontology

To clearly grasp our framework, it is useful to pre-

cise deﬁnitions of the concepts of model, metamodel,

megamodel and ontology to remove any ambiguity.

We ﬁt the deﬁnitions provided by Minsky and

ezivin for the concept of “model”. According to

the former author: “To an observer B, an object A

∗

is a model of an object A to the extent that B can

use A

∗

to answer questions that interest him about A”

(Minsky, 1965). As for Hill, B

ezivin and Gerb

e: “a

model is a simpliﬁcation of a system built with an in-

tended goal in mind. The model should be able to an-

swer questions in place of the actual system.” (Hill,

1996; B

ezivin and Gerb

e, 2001). Moreover, a func-

tion of representation mapping models in M1 with

the actual systems they represent in M0 is deﬁned in

ezivin and Gerb

e, 2001; B

ezivin, 2005), then re-

ﬁned in (Favre and Nguyen, 2005; Favre, 2006). This

function, often written “µ” and read “represents”, is

the essence of modeling; it is used from M1 to M0.

These deﬁnitions are sufﬁcient regarding the scope of

this paper.

We rely on Favre’s contributions regarding the

deﬁnition of the concept of “metamodel”. Accord-

ing to the latter, a metamodel is a model of language

(i.e., a grammar) through which downer-level models

could be expressed. To be more speciﬁc, a metamodel

could either be deﬁned as “a model of a language of

models” or “a model of a modeling language” (Favre

and Nguyen, 2005; Favre, 2006). B

ezivin and Favre

et al. both deﬁne a function of conformance, written

χ and read “conformsTo” or “conformantTo”. This

conformance relationship is used to link models to

metamodels from levels M1 to M2, metamodels to

the MOF from levels M2 to M3, and the MOF to it-

self at level M3 thanks to metacircularity (B

ezivin,

2005; Favre and Nguyen, 2005; Favre, 2006). The

conformance relationship maps a model with the cor-

responding grammar through which the model is ex-

pressed (Favre and Nguyen, 2005). Furthermore,

ezivin draw a parallel between the concept of meta-

model and the notion of ontology (B

ezivin, 2005).

Our framework requires the concept of “meg-

amodel”, ﬁrst introduced by B

ezivin to deﬁne “a

model which elements represent models, metamodels

and other global entities” according to (Favre, 2006).

We make use of this concept to express the major

relationships (i.e., representation, conformance, and

transformation) between systems, DT, models, meta-

models, and the MOF within the four metamodeling

layers. In addition, models could be transformed into

other models within these layers thanks to a common

metamodel; for that matter (Favre, 2006) introduced a

function of transformation, written τ and read “trans-

formedIn”.

Based on (Gruber, 1995; Guarino, 1998; Maed-

che, 2002; Gruber, 2008), we consider an ontology in

the computer science meaning. This supposed that, an

ontology is a set of concepts and axioms in relation-

ship with each other through semantic and taxonomic

relations. It represents a speciﬁc view of the world

(e.g., reality, particular domain) shared among a com-

munity. Concepts and relations in an ontology are ex-

plicitly expressed through a more or less formalized

language (e.g., OWL, UML) and axioms could be ex-

pressed with a mathematical or ﬁrst-order logic for-

malism.

3.3 Our Generic Proposal

3.3.1 Synoptic Megamodel

We illustrate our proposal of framework for produc-

ing a DT with MDE through our synoptic megamodel

presented in Figure 1.

In our megamodel, the DT is a digital system at

level M1 that represents (i.e., µ) the actual real-

world system at level M0. Stakeholders can inter-

act with the actual system through its virtual coun-

terpart (e.g., system management, monitoring, con-

trol/command). The virtual system can also ad-

dress other use-cases regarding the actual system,

such as predictions through statistics or simulations.

The digital representation is constituted of a set of

digital objects, instanceOf a digital library of ob-

jects (i.e., concepts describing the actual system). A

chosen object-oriented PSM is embedded in this li-

brary. The digital representation and the digital li-

brary from which it is instantiated are both contained

at level M1. As stated in (B

ezivin, 2005), there is a

programming conformance relationship between in-

stances and classes. The same apply here between the

digital representation and the digital library, however

it is not explicitly shown in the diagrams since only

metamodeling conformance relationships are drawn.

This library programmed in a given PSM is the

last product of chain of transformations following the

MDA: ﬁrst a CIM is deﬁned to formalize all the nec-

essary concepts and relationships related to the real-

world system. The production of the CIM is compa-

rable to the establishment of a domain ontology. This

CIM is then transformedIn (i.e., τ) into one or sev-

eral PIMs in order to satisfy development constraints

and choices. Each PIM is a software solution that is

then transformedIn into one or several PSMs that

are implementation models. CIM, PIMs, and PSMs

are all contained at level M1.

Linking Digital Twin Design and Ontologies with Model-Driven Engineering: Application to Railway Infrastructure

271

MOF

Modeling Language

Metamodel

Object-Oriented Programming

Language Metamodel

CIM

(Domain Model)

PIM

(Design Model)

PSM +

Library of objects

Digital Twin

System of interst

(Actual real-world system)

χ: conformsTo

χ: conformsTo χ: conformsTo

χ: conformsTo

instanceOf

µ: represents

τ: transformedIn τ: transformedIn

(generation)

Figure 1: Synoptic megamodel of our framework for producing a DT with MDA and the MOF.

The CIM and the PIMs conformsTo the meta-

model of the modeling languages they are expressed

with (e.g., OWL, UML). CIM and PIMs could be

expressed with the same or different modeling lan-

guages. The digital library of objects conformsTo the

metamodel of the object-oriented programming lan-

guage it is coded with. All these metamodels are con-

tained at level M2 and each of them conformsTo the

MOF at level M3.

3.3.2 Leveraging Ontologies to Formalize

Models

Following from the above, it clearly appears that the

very starting point of the production of the real-world

system digital representation, cornerstone of the DT,

is its corresponding domain model. The latter is ex-

pressed through as CIM using the MDA, as mentioned

previously. Then, one more question remains: how

could we produce this domain model? Since the do-

main is a formalization of a set of concepts in rela-

tionship describing the real-world, and more specif-

ically the scope concerned with the aforementioned

real-world system, it could be established such as an

ontology. This ontology would be a domain ontology,

according to Guarino’s well known classiﬁcation of

ontologies (Guarino, 1998; Maedche, 2002).

Furthermore, this domain ontology will require

to commit from an upper-level ontology (i.e., also

known as top-level ontology) to reuse, reﬁne, and

specialize really high-level concepts (Guarino, 1998;

Maedche, 2002). Several upper-level ontologies ex-

ist to date, one of the most popular is the Uniﬁed

Foundation Ontology (UFO) (Guizzardi, 2005). Oth-

ers could be mentioned such as Basic Formal Ontol-

ogy (BFO), Cyc, Descriptive Ontology for Linguistic

and Cognitive Engineering (DOLCE), General For-

mal Ontology (GFO), PROTo ONtology (PROTON),

Sowa’s Ontology, and Suggested Upper Merged On-

tology (SUMO). All these seven upper-level ontolo-

gies are compared in (Mascardi et al., 2007). We con-

jecture that the choice or the production of a top-level

ontology depends on ontological and epistemologi-

cal assumptions about what is reality and how do we

know what we know, including our knowledge about

reality. However, the scope of this paper is not to ad-

dress this question, but rather to open the discussion

on this subject.

3.4 Application at SNCF R

ESEAU

3.4.1 Applied Synoptic Megamodel

We present in Figure 2 the application of our proposal

of framework for producing a DT with MDE, previ-

ously presented in a generic way in Figure 1.

In our company, the actual system we aim at twin-

ing is the railway system. The Digital Twin, seen in

our framework as a shared up-to-date digital repre-

sentation of the railway system among the stakehold-

ers working at and with the company, embraces:

• Functional aspect of the infrastructure (e.g., re-

quirements, description of the French Railway

Network within schematics and/or Building Infor-

mation Modeling (BIM) mock-ups) considering

its complete lifecycle i.e., the as-designed phase

embracing the history of all the different versions

of each design study, the as-built phase consid-

ering the actual construction of the infrastructure

and the possible gaps with the as-designed re-

quirements, and ﬁnally the as-is phase focusing

DTO 2024 - Special Session on Ontologies for Digital Twin

272

MOF

UML Metamodel Java Metamodel

CIM

(Domain Model)

PIM: ARIANE

(Design Model)

PSM: Java +

Library of railway objects

Digital Twin

Railway system

χ: conformsTo

χ: conformsTo χ: conformsTo

χ: conformsTo

instanceOf

µ: represents

τ: transformedIn τ: transformedIn

(generation)

Figure 2: Synoptic megamodel of our framework for producing a DT, applied at SNCF R

ESEAU.

on the current conﬁguration of the infrastructure

in which it is operated. This functional aspect in-

clude the description of ﬁeld elements of the in-

frastructure e.g., tracks, tunnels, bridges, electri-

cal substations, catenaries, signaling items;

• Physical aspect of the infrastructure that is, actual

real-world ﬁeld elements deployed on the French

Railway Network considering their complete life-

cycle i.e., the as-built and the as-is phases. Spe-

ciﬁc components of deployed ﬁeld elements are

supervised for asset management and mainte-

nance purposes e.g., turnout frogs;

• Up-to-date measures and statuses related to the

physical infrastructure e.g., sensors controlling

the position of turnout points, sensors monitoring

the current consumption of point machines, sen-

sors ensuring rock-fall, ﬂood, and on-track hotbox

detectors. Each measurement and status is saved

through an history for further analysis;

• Railway capacity allocation of the rolling stock

e.g., scheduled train paths, real-time running train

circulations, working zones.

The digital representation is instantiated in reposito-

ries, thanks to our library of objects (i.e., classes de-

scribing railway concepts) in which the Java PSM is

embedded. Java is the main PSM used, since it has

been adopted as a coding standard in our company.

Therefore, the object-oriented programming meta-

model in our framework applied at SNCF R

ESEAU is

the Java metamodel. However, other object-oriented

languages could be used as well, such as C++ for sim-

ulation applications. In this case, the digital library

would be written in several object-oriented program-

ming languages, each one conforming to its corre-

sponding metamodel, all metamodels still conforming

to the MOF.

The digital library is produced based on our PIM

named “ARIANE” which is a software solution (i.e.,

a design model containing roughly a thousand classes

at the moment) based on the requirements of the CIM

(i.e., a railway domain model). All the concepts re-

lated to the railway system are contained in the CIM,

produced with a global, systemic, and ontological ap-

proach. “ARIANE” is a reference PIM for our com-

pany which has been chosen as input, especially its

topology package, for the establishment of the In-

ternational Union of Railways (UIC) standard called

“Rail System Model” (RSM) (Tane et al., 2022; Char-

train et al., 2024). At SNCF R

ESEAU, the CIM and

the PIM are both produced in UML, consequently

they both conform to the UML metamodel.

3.4.2 Digital Twin and Information System

At SNCF R

ESEAU, our DT is on the one hand, auto-

matically updated by the IoT (e.g., sensors) and, on

the other hand also manually updated by end-users

and data managers (Chartrain et al., 2024). It is our

reference source of information regarding the railway

system, guaranteeing both the unicity and the accu-

racy of data and therefore information (Issa et al.,

2024). As a result, the DT is a major component of

our Information System in the company.

We previously published our DT reference tech-

nical architecture in (Issa et al., 2024). It is achieved

with a Service-Oriented Architecture (SOA) and more

speciﬁcally a Representational State Transfer (REST)

Architecture, allowing end-users (e.g., stakeholders

working directly in our company or in partnership

with our company) to access whole or part of the

digital representation contained in the repositories

Linking Digital Twin Design and Ontologies with Model-Driven Engineering: Application to Railway Infrastructure

273

through RESTful web-services (Issa et al., 2024;

Chartrain et al., 2024). Software applications call

web-services and shape the information in the re-

quired format depending on the needs of each end-

user.

3.4.3 Example in Railways: Turnouts and

Measurements

A turnout, also called switch, is a railway ﬁeld ele-

ment made up of a set of points (i.e., movable ele-

ments of the turnout) and a frog (i.e., central cross-

ing piece of rail within the turnout). Turnouts en-

able the setting of a given direction at a local bifur-

cation within the route of a train. In this example we

are particularly interested in electrically commanded

and controlled turnouts, respectively by the way of

point machines and sensors that check the validity of

the commanded position after the movement of the

points.

The following aspects of turnouts are currently

stored within our DT repositories:

• A physical aspect including: (1) each identiﬁed

frog of each turnout within the scope of the French

Railway Network and (2) rails, sleepers, and bal-

last related to the track section within which the

turnout is embraced. These latter elements are

only instantiated at the scale of the track section

and not at the scale of the turnout;

• A functional aspect including: (1) the network

topology ensured by crossings and turnouts (2) a

precise description of each turnout with especially

its corresponding orientation (e.g., left, right, or

symmetric/Y-shaped), number of branches, num-

ber of points. This description is usually achieved

through track and signaling schematics. The in-

tegration of turnouts models contained in BIM

mock-ups within our DT is foreseen;

• Measurements related to the position of the points

and to the current consumption of point machines.

The execution of our framework regarding turnouts

and their related measurements (e.g., points position,

point machines current consumption) requires the es-

tablishment of a domain model for both of these no-

tions i.e., models of “turnout” and “measurement”.

These are components of a CIM, further transformed

into corresponding packages in our PIM that is, AR-

IANE at SNCF R

ESEAU. Based on the latter, cor-

responding Java classes related to the concepts of

“turnout” and “measurement” are generated. These

Java classes are then instantiated in our DT reposito-

ries with actual data about the functional and physical

aspects of turnouts, along with related measurements.

Figure 3 shows a simpliﬁed domain model re-

lated to a generic concept of “measurement” through

a UML class diagram. It is adapted for the case of

SNCF R

ESEAU in order to ﬁt the needs of our com-

pany. However, this model is not complete: it has

been reduced given the informative context provided

in this paper. In our approach, this generic model is

specialized for each context and measured item.

instrumentUsed

methodUsed

0..1

contextOfEvent

valueUnit

valueUncertainty

0..1

valueRefFrame

measurementAchieved

values

1..n

Event

MeasurementEvent

MeasuringInstrument

Method

ContextMeasurement

Value Unit

Uncertainty ReferenceFrame

Figure 3: Simpliﬁed generic measurement domain model.

Given that the access to our PIM ARIANE is re-

stricted for intellectual property reasons, we cannot

show a corresponding class diagram extracted from

our PIM. Nevertheless, we provide a link towards

RSM

, the UIC standard railway PIM that we have

previously mentioned above. In this latter model, a

model of the concepts of “turnout” and “measure-

ment” can be respectively found in the <<Domain>>

Track and the ObservationAndMeasure packages.

However, ARIANE is still more detailed at the mo-

ment and therefore our PIM related to the concept

of “measurement” is closer to the domain model pre-

sented in Figure 3 than the model presented in RSM.

To complete this example, we provide in Figure

4, an instance diagram (i.e., a UML objects diagram)

that shows how Java classes could be instantiated in

the DT repositories at SNCF R

ESEAU. In this ex-

ample, instances are related to the monitoring of a

point machine current consumption, when the points

are maneuvered from one side of the turnout to the

other by the point machines.

4 CONCLUSIONS

The concept of Digital Twin (DT) have ﬁrst appeared

in the early 2000’s. Since 2015, there has been a sig-

https://rsm.uic.org/doc/rsm/rsm-1-2/

DTO 2024 - Special Session on Ontologies for Digital Twin

274

instrumentUsed

contextOfEvent

PMinvolved

turnout

valueUnit

measurementAchieved

values

:PMcurrentMeasurementEvent

- endOfEventDateTime = 2019-02-27T05:19:09

::MeasurementEvent

::Event

+ dateTime = 2019-02-27T05:18:16

:PMcurrentContext

::Context

:PMcurrentSensor

::MeasuringInstrument

:PointMachine

:Turnout

:PMcurrentMeasure

::Measurement

- typeOfMeasurementInterval = temporal

- intervalValue = 0.1

:PMcurrentValue

::Value

- ranks = { 0, ..., 532 }

- values = { 3.6, ..., 3.1 }

:PMcurrentUnit

::Unit

- SIunit = ampere

- SIunitSymbol = A

PM means PointMachine.PM means PointMachine.

Figure 4: Simpliﬁed instantiation example illustrating the measurement of point machines current consumption.

niﬁcant growth in scientiﬁc literature on this topic,

and numerous concrete applications have emerged

in various sectors such as manufacturing, healthcare,

and transportation. DTs are promising tools to man-

age, monitor, control, simulate, and predict the be-

haviour of a real-world system.

In this paper, we ﬁrst provided an overview of

these concepts by studying their historical back-

grounds and conducted a state of the art of their deﬁni-

tions in literature. Based on our analysis, there is little

consensus regarding a generic, universal deﬁnition of

the DT concept across industry and academia. How-

ever, we emphasized that a clear deﬁnition is never-

theless required to guide the design and the produc-

tion of a DT. In addition, there are some frameworks,

methods, and standards to produce DTs but there is

currently no widely accepted generic or standardized

framework.

Furthermore, we underlined that a DT requires

models in order to be built, however among the ex-

isting frameworks, very few focus on the production

of model-driven DTs. Ontologies which are set of

concepts and axioms related to each other through se-

mantic and taxonomic relations, represent a speciﬁc

view of the world shared by a community. They of-

fer relevant solutions for analyzing domains, systems,

and products and formalize related computable mod-

els. The latter are of paramount importance to fur-

ther produce the digital representation of the system

or product to be digitally twined.

In this position paper, we argue that ontologies,

Model-Driven Engineering, and object-oriented prin-

ciples can be leveraged to produce model-driven Dig-

ital Twins. We propose a framework designed to as-

sist in the creation and development of DTs based on

these latter methods, hoping that it could help paving

the way towards a future standardized, generic frame-

work. We provide insights into the application of our

framework at SNCF R

ESEAU, the French Railway In-

frastructure Manager.

REFERENCES

Adamenko, D., Kunnen, S., Pluhnau, R., Loibl, A., and

Nagarajah, A. (2020). Review and comparison of

the methods of designing the Digital Twin. Procedia

CIRP, 91:27–32.

Aheleroff, S., Xu, X., Zhong, R. Y., and Lu, Y. (2021). Dig-

ital Twin as a Service (DTaaS) in Industry 4.0: An Ar-

chitecture Reference Model. Advanced Engineering

Informatics, 47:1–15.

Aßmann, U., Zschaler, S., and Wagner, G. (2006). Ontolo-

gies, Meta-models, and the Model-Driven Paradigm.

In Calero, C., Ruiz, F., and Piattini, M., editors, On-

tologies for Software Engineering and Software Tech-

nology, pages 249–273. Springer, Berlin, Heidelberg,

Allemagne.

Barricelli, B. R., Casiraghi, E., and Fogli, D. (2019). A

Survey on Digital Twin: Deﬁnitions, Characteristics,

Applications, and Design Implications. IEEE Access,

7:167653–167671.

Ben Sta, H., Ben Said, L., Gh

edira, K., Bigand, M., and

Bourey, J.-P. (2005). Cartographies of Ontology Con-

cepts. In Proceedings of the Seventh International

Conference on Enterprise Information Systems, pages

486–494, Miami, FL, USA. SciTePress - Science and

and Technology Publications.

ezivin, J. (2004). In Search of a Basic Principle for Model

Driven Engineering. The European Journal for the

Informatics Professional, Novatica, 5(2):21–24.

ezivin, J. (2005). On the Uniﬁcation Power of Models.

Softw Syst Model, 4(2):171–188.

ezivin, J. and Gerb

e, O. (2001). Towards a Precise Def-

inition of the OMG/MDA Framework. In Proceed-

Linking Digital Twin Design and Ontologies with Model-Driven Engineering: Application to Railway Infrastructure

275

ings of the 16th Annual International Conference on

Automated Software Engineering (ASE 2001), pages

273–280, San Diego, CA, USA. IEEE.

Chartrain, A., Dessagne, G., Haddad, N., and Hill, D. R.

(2024). Retrospective on the Digital Twin concept and

perspectives for railways: the case of SNCF R

eseau.

In 2

eme Congr

es Annuel de la SAGIP (SAGIP ’24),

Lyon, France. Soci

e d’Automatique, de G

enie In-

dustriel et de Productique (SAGIP).

Dahl, O.-J. and Nygaard, K. (1966). SIMULA: An

ALGOL-based simulation language. Comm. ACM,

9(9):671–678.

De Donato, L., Dirnfeld Turocy, R., Somma, A., De Bene-

dictis, A., Flammini, F., Marrone, S., Samanazari, M.,

and Vittorini, V. (2023). Towards AI-assisted digital

twins for smart railways: Preliminary guideline and

reference architecture. Journal of Reliable Intelligent

Environments, 9:303–317.

Favre, J.-M. (2006). Megamodelling and Etymology.

In Transformation Techniques in Software Engineer-

ing, volume 5161 of Dagstuhl Seminar Proceed-

ings, Wadern, Germany. Schloss Dagstuhl-Leibniz-

Zentrum f

ur Informatik.

Favre, J.-M. and Nguyen, T. (2005). Towards a Megamodel

to Model Software Evolution Through Transforma-

tions. Electr. Notes Theor. Comput. Sci., 127:59–74.

Glaessgen, E. and Stargel, D. (2012). The Digital Twin

Paradigm for Future NASA and U.S. Air Force Ve-

hicles. In 53rd Structures, Structural Dynamics, and

Materials Conference: Special Session on the Digital

Twin, pages 1–14, Honolulu, Hawaii. American Insti-

tute of Aeronautics and Astronautics.

Grieves, M. (2015). Digital Twin: Manufacturing Excel-

lence through Virtual Factory Replication. White Pa-

per, 1:1–7.

Grieves, M. and Vickers, J. (2017). Digital Twin: Mitigat-

ing Unpredictable, Undesirable Emergent Behavior in

Complex Systems. In Kahlen, F.-J., Flumerfelt, S.,

and Alves, A., editors, Transdisciplinary Perspectives

on Complex Systems, pages 85–113. Springer Interna-

tional Publishing, Cham.

Gruber, T. (2008). Deﬁnition of Ontology.

https://tomgruber.org/writing/deﬁnition-of-ontology.

Gruber, T. R. (1995). Toward Principles for the Design

of Ontologies Used for Knowledge Sharing? Inter-

national Journal of Human-Computer Studies, 43(5-

6):907–928.

Guarino, N. (1998). Formal Ontology and Information Sys-

tems. In Formal Ontology in Information Systems.

Proceedings of the First International Conference

(FOIS ’98), Frontiers in Artiﬁcial Intelligence and Ap-

plications, pages 3–15, Trento, Italy. IOS Press.

Guizzardi, G. (2005). Ontological Foundations for Struc-

tural Conceptual Models. PhD thesis, University of

Twente, Enschede, The Netherlands.

Haße, H., van der Valk, H., M

oller, F., and Otto, B. (2022).

Design Principles for Shared Digital Twins in Dis-

tributed Systems. Bus Inf Syst Eng, 64(6):751–772.

Hill, D. R. C. (1996). Object-Oriented Analysis and Simu-

lation. Addison Wesley, Harlow, England ; Reading,

MA, USA, 1st edition.

Issa, M., Chartrain, A., Viguier, F., Landes, B., Dessagne,

G., Haddad, N., and Hill, D. R. (2024). Railway sys-

tem Digital Twin: A tool for extended enterprises to

perform multimodal transportation in a decarboniza-

tion context. In Transport Research Arena 2024 (TRA

’24), Dublin, Ireland.

Kritzinger, W., Karner, M., Traar, G., Henjes, J., and Sihn,

W. (2018). Digital Twin in manufacturing: A cat-

egorical literature review and classiﬁcation. IFAC-

PapersOnLine, 51(11):1016–1022.

Madni, A. M., Madni, C. C., and Lucero, S. D. (2019).

Leveraging Digital Twin Technology in Model-Based

Systems Engineering. Systems, 7(1):1–13.

Maedche, A. (2002). Ontology — Deﬁnition & Overview.

In Maedche, A., editor, Ontology Learning for the Se-

mantic Web, volume 665 of The Kluwer International

Series in Engineering and Computer Science, pages

11–27. Springer, Boston, MA, USA, 1 edition.

Mascardi, V., Cord

ı, V., and Rosso, P. (2007). A Com-

parison of Upper Ontologies. In Dagli Oggetti Agli

Agenti. Agents and Industry: Technological Applica-

tions of Software Agents, pages 55–64, Genova, Italy.

Seneca Edizioni.

Minsky, M. L. (1965). Matter, Mind and Models. In Proeed-

ings of International Federation of Information Pro-

cessing Congress 1965, volume 1, pages 45–49, New-

York City, NY, USA.

Segovia, M. and Garcia-Alfaro, J. (2022). Design, Mod-

eling and Implementation of Digital Twins. Sensors,

22(14):1–30.

Semeraro, C., Lezoche, M., Panetto, H., and Dassisti, M.

(2021). Digital twin paradigm: A systematic literature

review. Computers in Industry, 130:1–23.

Tane, P., Dessagne, G., Janssen, B., and Magnien, A.

(2022). The case for a federated digital model of the

rail system. Global Railway Review.

Tao, F. and Qi, Q. (2019). Make more digital twins. Nature,

573(7775):490–491.

Tao, F., Xiao, B., Qi, Q., Cheng, J., and Ji, P. (2022). Digital

twin modeling. Journal of Manufacturing Systems,

64:372–389.

Valiente, M.-C., Vicente-Chicote, C., and Rodr

ıguez, D.

(2011). An Ontology-Based and Model-Driven Ap-

proach for Designing IT Service Management Sys-

tems. International Journal of Service Science, Man-

agement, Engineering, and Technology, 2(2):65–81.

VanDerHorn, E. and Mahadevan, S. (2021). Digital Twin:

Generalization, characterization and implementation.

Decision Support Systems, 145:1–11.

Wright, L. and Davidson, S. (2020). How to tell the differ-

ence between a model and a digital twin. Adv. Model.

and Simul. in Eng. Sci., 7:1–13.

Zhang, L., Zhou, L., and Horn, B. K. P. (2021). Building a

right digital twin with model engineering. Journal of

Manufacturing Systems, 59:151–164.

DTO 2024 - Special Session on Ontologies for Digital Twin

276