=Paper=
{{Paper
|id=Vol-3618/forum_paper_21
|storemode=property
|title=Using goal modeling for defining digital twins in industry automation
|pdfUrl=https://ceur-ws.org/Vol-3618/forum_paper_21.pdf
|volume=Vol-3618
|authors=Jeshwitha Jesus Raja,Meenakshi Manjunath,Philipp Kranz,Fabian Schirmer,Marian Daun
|dblpUrl=https://dblp.org/rec/conf/er/RajaMKSD23
}}
==Using goal modeling for defining digital twins in industry automation==
https://ceur-ws.org/Vol-3618/forum_paper_21.pdf
Using goal modeling for defining digital twins in
industry automation
Jeshwitha Jesus Raja, Meenakshi Manjunath, Philipp Kranz, Fabian Schirmer and
Marian Daun
Center for Robotics, Technical University of Applied Sciences Würzburg-Schweinfurt, Schweinfurt, Germany
Abstract
Goal modeling is commonly proposed and used for early requirements engineering, allowing easy
documentation of high-level requirements and their interrelations. In recent years, goal models have
often been suggested to support the definition and analysis of cyber-physical systems, also from the
industry automation domain. For the future of manufacturing, digital twins play a vital role. The digital
twin allows for in-depth analysis of the requirements and processes of the factory and supports action
planning for new tasks. In this paper, we investigate the use of the goal-oriented requirement language
(GRL) to define goal models that support the digital twin in industry automation.
Keywords
Goal Modeling, GRL, Digital Twin, Industry Automation
1. Introduction
Digital twins are virtual representations of physical systems [1]. In smart manufacturing, digital
twins are used to monitor and analyze the overall state of the factory and its production processes.
Therefore, the digital twin helps in runtime adaptation planning, due to new production orders
or due to the identification of unforeseen events or the occurrence of errors during process
execution. In industry automation, digital twins, for instance, allow monitoring production
processes, calculating optimal factory allocation, or predicting manufacturing outcomes and
potential defects (cf., e.g., [2, 3]).
Among others, the huge potential of digital twins is seen in supporting safety assurance at
run-time, exploring problem spaces to find optimal or near optimal solutions, foster prediction
of run-time properties [4, 3]. One of the primary challenges encountered with digital twins
is its systematic definition, particularly considering generalization and managing large-scale
operations [5, 6]. A digital twin doesn’t achieve immediate perfection; it necessitates the
accumulation of extensive data sets before it can autonomously make accurate predictions. This
could take quite some time and these predictions also require a variety of data (positive and
negative), to build the digital twin. Therefore, currently, digital twins are mostly used as research
ER2023: Companion Proceedings of the 42nd International Conference on Conceptual Modeling: ER Forum, 7th SCME,
Project Exhibitions, Posters and Demos, and Doctoral Consortium, November 06-09, 2023, Lisbon, Portugal
Envelope-Open jeshwitha.jesusraja@study.thws.de (J. Jesus Raja); meenakshi.manjunath@study.thws.de (M. Manjunath);
philipp.kranz@thws.de (P. Kranz); fabian.schirmer@thws.de (F. Schirmer); marian.daun@thws.de (M. Daun)
Orcid 0009-0008-7886-7081 (J. Jesus Raja); 0009-0005-6421-1450 (M. Manjunath); 0000-0002-1057-4273 (P. Kranz);
0000-0002-7032-8242 (F. Schirmer); 0000-0002-9156-9731 (M. Daun)
© 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
prototypes developed by data scientists in a trial and error approach. This is exacerbated as in the
industry automation domain no strong emphasis is given to systematic software development
but more to mechanical engineering aspects. To support the systematic development of digital
twins in manufacturing scenarios, goal models [7, 8] might be a promising solution as it allows
early definition and analysis of the system and its digital twin at moderate effort [9, 10].
This paper contributes an investigation to a) evaluate whether goal modeling can be ben-
eficially applied to the development of digital twins for manufacturing use cases in industry
automation and b) identify necessities to be taken into account for the definition of a compre-
hensive goal modeling approach for the specification of digital twins in industry automation use
cases. Thus, this research is part of a larger research agenda for the model-based development
of digital twins in smart manufacturing.
The paper is outlined as follows. Section 2 gives an overview of the related work, Section 3
introduces a running example of a smart factory, which is used to illustrate the approach and
showcase the benefits of using a goal model as a digital twin. Section 4 introduces the approach
under investigation to derive the goal models of the system and the digital twin. Section, 5
discusses findings from the applying goal modeling to specify a digital twin for our industry
use case. Finally, Section 6 concludes the paper.
2. Related work
Research on goal modeling often focuses on the development of profiles to accustom the needs
of specific application domains and systems. A comprehensive review of the goal modeling
research field is given by Horkoff et al. in [11].
Recently, the systems and components under development or investigation are often modeled
as actors rather than the stakeholders [12]. This, is particularly the case for describing modern
system types. Therefore, various iStar profiles have been proposed for the development of
cyber-physical systems (CPS, e.g., [13, 14]) or systems-of-systems (SoS, e.g., [15, 16]). The
idea behind these approaches is viewing the CPS as interacting, collaborating, or competing
actors. Thus, the focus of investigation moves from the original questions ’Can we fulfill all
requirements of the involved stakeholders?’ and ’Which stakeholders are in conflict with one
another?’ to ’Are different systems compatible?’ and ’Will they collaborate correctly?’ [17]. In
this sense, correctness and compatibility is assumed when two systems can work together
and the desired goals of both systems or of an overall system network can be fulfilled. The
goal modeling approach in this paper, is oriented on this idea, as we also describe CPS that
collaborate in a SoS.
For the use of goal models at runtime, various approaches exist. Most famously to mention
is Tropos4AS [18] and extensions thereof, such as a recent approach by Morgan et al. [19].
These approaches extend the goal modeling language (most commonly iStar) with explicit
context modeling [20]. This allows, among others, monitoring and proposing adaptations due
to recognized context changes. Other approaches, e.g., the approach by Dalpiaz et al. [21],
focus on the interactions between different actors, and define adaptation strategies for these. In
contrast, our approach does not focus on the context to be monitored, but aims at providing
support for monitoring the system (i.e. for developing a digital twin).
Monteiro et al. [22] proposed the use of goal models to develop a digital twin for vertical
farming. They showed that the use of goal modeling has benefits for developing a digital twin.
In our case, we want to systematically evolve the goal model of the original system, rather than
create a digital twin goal model from scratch. In addition, contrasting the approach of Monteiro
et al., we want to stick as close as possible to standard goal modeling languages (i.e. GRL and
iStar) to reduce the number of modeling elements needed for describing the digital twin.
3. Case example: Smart factory
A smart factory (also referred to as U–Factory (ubiquitous factory), or factory of things) is a
highly digitized and networked production facility [23, 24], whose concept is based on two
major ideas:
• Smart manufacturing aims at completely monitoring the factory and its production pro-
cesses for deriving ubiquitously valuable knowledge created/managed by the factory (or
its digital twin). This allows, among others, for optimization of production processes[25],
fast error response [3], and long term improvements of the smart factory.
• Flexible production aims at defining production processes based on the product. Thus, a
smart factory is designed to develop a plethora of different products that are unknown at
design time of the smart factory [26]. In traditional industry automation, the product is
defined first (i.e. as CAD model), and then the production process (i.e. the bill of process)
is defined. The production process specifies how this product shall be produced. Finally,
a factory is built or re-organized in accordance with the production process. In contrast,
in the smart factory, the factory is build first. The production process is defined based on
the product specification and the factory’s available capabilities [27].
In our research, we use an assembly line as a case example. The assembly line consists of a
set of collaborative robots equipped for different welding purposes. Along this setup, there exist
multiple monitoring systems needed for the digital twin, which are attached to every cobot
station that analyze the execution of the respective process step. There are three important
things to monitor: (1) the position of the cobot and the work piece, (2) safety aspects, (3) the
outcome of the process.
Currently, research on digital twins mostly focuses on the later phases. Therefore, research
questions are often data-centric in nature to derive precise predictions to trigger adaptations.
As various reasoning approaches have been proposed for goal models, goal models might also
be an interesting mean to support runtime analysis of the digital twin. However, in this paper,
we want to focus on the preconditions, i.e. whether manufacturing systems as well as their
digital twins can benefit from using goal models for their specification. We will investigate
whether modeling specific objectives and requirements of the respective welding processes
is possible and provides support for the development of digital twins. For the case example.
we specified distinct goal models for the different participating cobots and their monitoring
systems, and their interrelations.
4. Application to the case example
4.1. Overview over the goal modeling approach under investigation
In previous work, we have already shown that iStar [12] and the GRL [28] are good modeling
languages to describe collaborative CPS in the industry automation domain [29] and proposed
a GRL compliant iStar extension [14] to foster a systematic yet structured approach to specify
collaborative CPS. In this paper, we use this approach to also specify the digital twin, based
upon the specification of the system. Therefore, a two-step modeling approach is used. First,
the smart factory and the individual cyber-physical production and transport systems of the
smart factory are defined using a GRL goal model. Based on the goal model, the digital twin is
defined. Therefore, the initial goal model shows the working steps of one of the cobots and the
outputs it should achieve at the end of the process. Note that this simplified process illustrates
how the digital twin can be developed using goal modeling. However, for a practical application
- as we will discuss in our findings - a more nuanced process including iterations is needed, as
the design of the digital twin influences the system design and vice versa.
4.2. Modeling the physical twin
We need to ensure that the goal model can be properly connected to the real world. Important
for linking the goal model to the real world are tasks. According to the Z.151 standard [28],
tasks are defined as “a set of instructions that are executed by a system in order to accomplish a
specific goal or objective.” They emphasize on the functional nature of tasks and highlights their
purpose in achieving specific outcomes. Thus, tasks are an abstract description of the behavior
of the system. This in turn can be monitored, and its fulfillment can be determined. Therefore,
we restrict the leafs of the goal graph to be tasks, as this allows for proper monitoring.
COBOT
Weld parts
together
AND
Position
Welding
calculator
AND AND
Calibrate cobot Connect parts
position w/ pressure
Check position
Heat-up spot
of part
Figure 1: Goal model of the spot arc welding cobot
Figure 1 shows this restrictive modeling approach for the smart factory. Looking into how
the assembly line is designed, it is visible that the whole process is divided into individual tasks
and each end-effector completes a particular task. In Figure 1, we focus on just one of these
cobots conducting spot welding tasks to assemble the car’s doors to the chassis.
The goal model shows all tasks the cobot needs to execute in order to weld the door to the
chassis. Therefore, the cobot needs to calculate the position, and then it needs to move the spot
welder to that exact position and weld the door to the car. The sub-tasks “check position of
part” and “calibrate cobot position” are essential, first the position of the part in relation to the
position of the cobot arm is determined and then subsequently the cobot needs to be calibrated
accordingly. The welding consists of two aspects, which are also specified as tasks. The cobot
needs to pre-heat the welding spot and thereafter put pressure on the materials to actually weld
the door into the chassis.
4.3. Modeling the digital twin
Next, we define the goal model for the digital twin. Therefore, we extend the goal model of the
cobot (Figure 1) with aspects needed for properly executing the digital twin. The resulting goal
model for the digital twin is shown in Figure 2. Most notably, we added monitoring tasks to
monitor system execution. We, explicitly, distinguish these monitoring tasks belonging to the
digital twin, from the execution tasks belonging to the system.
COBOT
Weld parts
together
AND
Position
Welding
calculator
AND AND
Calibrate cobot Connect parts
position w/ pressure
Check position
Heat-up spot
of part
Camera 1
Check
position Monitor
Monitoring Monitor position Monitor welding
temperature
System
Check welding
Camera 2 product
AND
Ensure successful completion
Temperature of welding
sensor
Figure 2: Goal model of the digital twin, including the monitoring system
There are also multiple sensors and cameras located to monitor the system during run-time.
There are certain possibilities where system malfunctions can happen. These further defined
monitoring tasks are dependent on the process tasks of the cobot. This allows us to express,
which process task is monitored by which monitoring task. In addition, we can define further
monitoring tasks that are not directly related to the process tasks. For instance, we can define a
monitoring task to monitor the work product after all the process tasks have been successfully
executed.
5. Findings from application
We evaluated the use of GRL goal models to model industrial production systems and to define
the digital twin using goal models using the case example outlined in Section 3. As a first result,
we can state that GRL is appropriate to specify not only the robotic systems in manufacturing
use cases, but also the digital twin. Furthermore, we found that building the digital twin goal
model upon the system’s goal model helps in specifying the digital twin. Particularly, this offers
support to identify tasks of the digital twin and to directly link them to the real world tasks and
processes, aiding in the definition of appropriate monitoring and data gathering strategies.
For our second question, i.e. identifying specificities and requirements for a goal modeling
approach to specify digital twin in industry automation, we recognized modeling patterns in
the system model and the digital twin model that can aid in defining a structured approach.
For the modeling of cobots in the production process, we have – at least in our industrial
case example – always two major branches:
• The calibration where the cobot identifies its position, the workpieces position, and
checks for safety concerns (e.g., humans or other cobots are blocking the work space).
• The execution of the production step. Here, we typically see some pre-processing (e.g.,
heating up the mounted tool) and the actual processing (e.g., spot welding).
For the monitoring system, we also see patterns. First, we mirror the production tasks with
monitoring tasks, second we need independent visual (e.g., a second camera as we cannot rely on
just one camera) and other sensors as well as a final monitoring task that does not monitor the
process execution but the result of the process. We checked this with other industry examples
from automated production processes and discussed this with industry experts.
In addition, we found two specifics that should be addressed with a modeling approach. First,
there is a need for specifying the production processes and linking these to the monitoring
process of the digital twin. While the process steps can be modelled using tasks, the goal model
does not specify any kind of execution order. However, for defining the tasks of the digital twin
this is crucial as the identification of defects not only relies on the current step monitored but also
on the outcome of previous production steps. Second, there is a need to specify different kinds
of dependencies. It is important to differentiate between a dependency from the monitoring
system needed by the digital twin to the system and between other dependencies i.e. between
different cobots or even between monitoring system and cobot, which are not grounded in the
digital twin, but stem, e.g., from usage of shared resources.
6. Conclusion and future work
In this paper, we investigated the use of GRL goal models to develop digital twins in the industry
automation domain. GRL goal models allow for easy documentation of high-level requirements
and their interrelations. We have shown that GRL goal models are applicable to model not
only the cyber-physical production systems within a factory, but also the needed monitoring
systems belonging to the digital twin. This way, the digital twin can be developed closely
linked to the individual goals of the systems to be developed. In consequence, it can be ensured
that every major production step – which is defined as task contributing to a major goal of a
production system – is monitorable and will be monitored by the digital twin. In addition, we
identified requirements to be considered when designing a goal modeling approach for digital
twins in smart manufacturing, as well as some useful modeling patterns. Thus, for the future,
involving fine-grained patterns in the goal-based definition of robotic production systems and
their digital twin can help in easier definition and also support the analysis by pre-configured
steps. Furthermore, for future work we plan to investigate the use of digital twin goal models
as runtime as the existing reasoning approaches can provide support to runtime analysis and
thereby the definition of monitoring and adaptation algorithms of the digital twin.
Acknowledgments
This research was partly funded by the Bayerische Forschungsstiftung under grant AZ-1512-21.
References
[1] C. Koulamas, A. Kalogeras, Cyber-physical systems and digital twins in the industrial
internet of things [cyber-physical systems], Computer 51 (2018) 95–98.
[2] I. Rojek, D. Mikołajewski, E. Dostatni, Digital twins in product lifecycle for sustainability
in manufacturing and maintenance, Applied Sciences 11 (2020) 31.
[3] F. Tao, Q. Qi, L. Wang, A. Nee, Digital twins and cyber–physical systems toward smart
manufacturing and industry 4.0, Engineering 5 (2019) 653–661.
[4] C. Liu, P. Jiang, W. Jiang, Web-based digital twin modeling and remote control of cyber-
physical production systems, Robotics and computer-integrated manufacturing 64 (2020).
[5] M. Attaran, B. G. Celik, Digital twin: Benefits, use cases, challenges, and opportunities,
Decision Analytics Journal (2023) 100165.
[6] A. Fuller, Z. Fan, C. Day, C. Barlow, Digital twin: Enabling technologies, challenges and
open research, IEEE access 8 (2020) 108952–108971.
[7] A. Van Lamsweerde, Goal-oriented requirements engineering: A guided tour, in: 5th ieee
Int. Symp. on requirements engineering, IEEE, 2001, pp. 249–262.
[8] E. S. Yu, Towards modelling and reasoning support for early-phase requirements engi-
neering, in: 3rd IEEE Int. Symp. on Requirements Engineering, IEEE, 1997, pp. 226–235.
[9] J. Horkoff, E. Yu, Interactive goal model analysis for early requirements engineering,
Requirements Engineering 21 (2016) 29–61.
[10] J. Brings, M. Daun, T. Bandyszak, V. Stricker, T. Weyer, E. Mirzaei, M. Neumann, J. S.
Zernickel, Model-based documentation of dynamicity constraints for collaborative cyber-
physical system architectures, Journal of systems architecture 97 (2019) 153–167.
[11] J. Horkoff, F. B. Aydemir, E. Cardoso, T. Li, A. Maté, E. Paja, M. Salnitri, L. Piras, J. My-
lopoulos, P. Giorgini, Goal-oriented requirements engineering: an extended systematic
mapping study, Requirements engineering 24 (2019) 133–160.
[12] F. Dalpiaz, X. Franch, J. Horkoff, istar 2.0 language guide, arXiv:1605.07767 (2016).
[13] C. Cares, S. Sepúlveda, C. Navarro, Agent-oriented engineering for cyber-physical systems,
in: Information Technology and Systems: ICITS 2019, Springer, 2019, pp. 93–102.
[14] M. Daun, J. Brings, L. Krajinski, V. Stenkova, T. Bandyszak, A grl-compliant istar extension
for collaborative cyber-physical systems, Requirements Engineering 26 (2021) 325–370.
[15] E. Cavalcante, T. Batista, N. Bencomo, P. Sawyer, Revisiting goal-oriented models for
self-aware systems-of-systems, in: Int. Conf. Autonomic Comp., IEEE, 2015, pp. 231–234.
[16] G. A. Lewis, E. Morris, P. Place, S. Simanta, D. B. Smith, Requirements engineering for
systems of systems, in: 2009 3rd annual IEEE systems Conf., IEEE, 2009, pp. 247–252.
[17] J. Brings, M. Daun, T. Weyer, K. Pohl, Goal-based configuration analysis for networks of
collaborative cyber-physical systems, in: ACM Symp. Applied Comp., 2020, pp. 1387–1396.
[18] M. Morandini, L. Penserini, A. Perini, A. Marchetto, Engineering requirements for adaptive
systems, Requirements Engineering 22 (2017) 77–103.
[19] R. Morgan, S. Pulawski, M. Selway, W. Mayer, G. Grossmann, M. Stumptner, A. Ghose,
R. Kyprianou, Modeling rates of change and aggregations in runtime goal models, in:
Conceptual Modeling: 41st Int. Conf., ER 2022, Springer, 2022, pp. 397–412.
[20] R. Ali, F. Dalpiaz, P. Giorgini, A goal-based framework for contextual requirements
modeling and analysis, Requirements engineering 15 (2010) 439–458.
[21] F. Dalpiaz, A. K. Chopra, P. Giorgini, J. Mylopoulos, Adaptation in open systems: Giving
interaction its rightful place, in: Int. Conf. Conceptual Modeling, Springer, 2010, pp. 31–45.
[22] J. Monteiro, J. Barata, M. Veloso, L. Veloso, J. Nunes, A scalable digital twin for vertical
farming, Journal of Ambient Intelligence and Humanized Computing (2022) 1–16.
[23] J.-S. Yoon, S.-J. Shin, S.-H. Suh, A conceptual framework for the ubiquitous factory, Int.
Journal of Production Research 50 (2012) 2174–2189.
[24] D. Lucke, C. Constantinescu, E. Westkämper, Smart factory-a step towards the next
generation of manufacturing, in: Manufacturing Systems and Technologies for the New
Frontier: The 41 st CIRP Conf. on Manufacturing Systems, Springer, 2008, pp. 115–118.
[25] T. H.-J. Uhlemann, C. Lehmann, R. Steinhilper, The digital twin: Realizing the cyber-
physical production system for industry 4.0, Procedia Cirp 61 (2017) 335–340.
[26] E. Hozdić, Smart factory for industry 4.0: A review, Int. Journal of Modern Manufacturing
Technologies 7 (2015) 28–35.
[27] M. Daun, J. Brings, P. A. Obe, S. Weiß, B. Böhm, S. Unverdorben, Using view-based
architecture descriptions to aid in automated runtime planning for a smart factory, in:
IEEE Int. Conf. on Software Architecture Companion (ICSA-C), 2019, pp. 202–209.
[28] ITU Int. Telecommunication Union, Recommendation ITU-T Z.151: User Requirements
Notation (URN), Technical Report, 2018.
[29] M. Daun, V. Stenkova, L. Krajinski, J. Brings, T. Bandyszak, T. Weyer, Goal modeling for
collaborative groups of cyber-physical systems with grl, in: ACM Symp. Applied Comp.,
2019, pp. 1600–1609.