A Framework for Model-Driven Digital Twin Engineering

Slides from my PhD Defense at Johannes Kepler University, held on Janurary 10, 2025.
The full thesis is available here: https://epub.jku.at/urn/urn:nbn:at:at-ubl:1-83896

1. Christian Doppler Laboratory for Model-Integrated Smart Production
Institute of Business Informatics –Software Engineering
Johannes Kepler University Linz
AltenbergerStraße69, Science Park 3
4040 Linz
Christian Doppler Laboratory for Model-Integrated Smart Production
CDL-MINT
A Framework for Model-Driven Digital Twin Engineering
PhD Defense –Daniel Lehner

2. What is a Digital Twin (DT)?
2Kritzinger, W., Karner, M., Traar, G., Henjes, J., & Sihn, W. (2018). Digital Twin in manufacturing: A categorical literature review
and classification.IFAC-PapersOnline,51(11), 1016-1022.
Physical Object
Digital Model Digital Shadow Digital Twin (DT)
Physical Object
Physical Object named
Physical Twin (PT)
Manual data flow
Automated data flow
Digital Model Digital Shadow Digital Twin
Legend

3. Conceptualization of DTs
3
DT System
Physical Twin (PT) DT Platform
communicates
DT Services
Simulation
«simulates»
«uses»
«uses»
Car
-manufacturer: string
-batteryLevel: int
-temperature: int
-toggleHeating()
-toggleCooling()
-toogleEnergyMode()
«StateMachine»
«State»
HeatingOn
«State»
HeatingOff
toggle
Heating()
toggle
Heating()
Interface „Car“ (DTMI: „dtdl.v2.com.carcompany.car:1“):
Property „manufacturer“ (String, isWritable=false)
Telemetry „batteryLevel“ (float)
Telemetry „temperature“ (int, type=„Temperature“,
unit=„Celsius“)
Command „toggleHeating“
Command „toggleCooling“
Command „toggleEnergyMode“
hosts
Digital Twin (DT)
DT ModelEngineering Model
«represents»
Engineering
Modeling Language
DT Modeling
Language
«c2» «c2»
«represents»
«is_a» «is_a»
Component
Legend
Dependency
Artefact
Inheritance

4. Model-Driven Engineering (MDE) of DTs
4
DT System
Physical Twin (PT) DT Platform
communicates
DT Services
Simulation
«simulates»
«uses»
«uses»
hosts
Digital Twin (DT)
DT ModelEngineering Model
«represents»
Engineering
Modeling Language
DT Modeling
Language
«c2» «c2»
«represents»
«abstract»
Model
Transformation
Engine
«input» «output»
Generation
Engine
«input»
«output»
Component
Legend
Dependency
Artefact
Inheritance

5. Problem 3: No support for
managing variability of
different DT Systems
DT System 1
Problems when developing DT Systems
5
Problem 2: Redundancy when
connecting DT services to
both PT and simulations
Problem 1: Redundancy
when specifying DT models
in addition to
engineering models.
DT System 2
Simulation
Physical Twin (PT)
DT Platform
DT ModelEngineering Model
Engineering Modeling
Language
DT Modeling
Language
«represents»
«represents»
«conformsTo» «conformsTo»
DT Service 1
DT Service 3
Simulation
DT Service 1 DT Service 2
DT Platform
«communicates»
«communicates»
«uses» «uses» «uses»
«uses»
«uses» «uses»
«uses»

6. Methodological Approach: Design Science Research
6Wieringa, R. J. (2014).Design science methodology for information systems and software engineering. Springer.
(1) Problem investigation:
What are problems when engineering DT Systems
using existing DT platforms?
(3) Research Method
Structured Review, Exemplar
(4) Available Treatments
What are existing MDE approaches for Digital Twins
in the literature?
(7) Research Method
Systematic Mapping Study using
Taxonomy
(8) Requirements satifsfaction
Problems
•Redundancy when specifying DT models in
addition to engineering models
•Redundancy when connecting DT services to both
PT and simulations
•No support for managing variability of different DT
systems
(11) Validation
Comparing requirements satisfaction + required effort
-New treatment
-Available treatments
(10) Prototype design and implementation
Contribution 1: Transformations between Engineering and DT Models
Contribution 2: Common interface to connect services to DT and simulations
Contribution 3: Method for template-based DT system modeling
(12) Research Method
Case Study Methodology
(9) New Treatment
DT++ Framework
Design Problem: How to reduce the effort for developing and maintaining DT systems by employing MDE techniques to automate the (i) setup and
maintenance of DTs running in DT platforms, (ii) connection of DT services to the PT and its simulations, and (iii) integration of DT platforms, simulations, and
DT services into DT systems, to make it easier for engineers to develop such DT systems.
(6) Design of Taxonomy
•Goals + Requirements for
Taxonomy
•Available Taxonomies?
•Design Taxonomy for MDE
•Design Taxonomy for DT
•Interlink both Taxonomies
•Requirements satisfaction?
(5) Design a Taxonomy of MDE for
DTs
(2) Knowledge Questions
Capabilities of DT platforms?
Modeling languages used by DT
platforms?

7. Problem Investigation
7
DT Exemplar for Air Quality Measurement [1]
•Physical Twin
•Digital Twin in Azure Digital Twins
•DT services for visualization and prediction
Structured review of existing DT platforms [2]
•Features extracted from ISO 25010
•Framework for adding new platform evaluations available
Review of modeling capabilities of DT platforms [3]
•Reverse-engineered modeling languages
•Common features
•Comparison with UML
•UML profile for DT modeling
[1] Govindasamy, H. S., Jayaraman, R., Taspinar, B., Lehner, D., & Wimmer, M. (2021). Air quality management: An exemplar for model-driven digital twin engineering. MODELS Companion .
[2] Lehner, D., Pfeiffer, J., Tinsel, E. F., Strljic, M. M., Sint, S., Vierhauser, M., ... & Wimmer, M. (2021). Digital twin platforms: requirements, capabilities, and future prospects.
IEEE Software, 39(2).
[3] Pfeiffer, J., Lehner, D., Wortmann, A., & Wimmer, M. (2022). Modeling capabilities of digital twin platforms-old wine in newbottles?. Journal of Object Technologies, 21(3).

8. Systematic Mapping Study
8Lehner, D., Zhang, J., Pfeifer, J., Splettstößer, A., Sint, S., Wortmann, A., & Wimmer, M. Model-Driven Engineering for Digital Twins: A
Systematic Mapping Study. Journal of Software and Systems Modeling (accepted for publication)
(“Digital Twin” OR “Digital Twins”)
AND
(Model-Driven OR “Model Driven”)
IEEE
73
ACM
64
Scopus
182
Inclusion
Criteria
Exclusion
Criteria
Manual screening removes
37 irrelevant publications
Detailed reviewing removes 86 publications
Snowballing adds another 1 publication
Forward
Snowballing
Backward
Snowballing
WoS
122
Automated processing removes
252 duplicates
Increasing research interest Few evaluation research

9. Methodological Overview
9
Reduce effort for engineering DT systems by employing MDE techniques
Redundant work when specifying DT models
in addition to engineering models
Redundant work when connecting DT
services to both PT and simulations
Lack of dedicated support for
managing the evolution of DT systems
Problems
Research Goal

10. Methodological Overview
Reduce effort for engineering DT systems by employing MDE techniques
Redundant work when specifying DT models
in addition to engineering models
Redundant work when connecting DT
services to both PT and simulations
Lack of dedicated support for
managing the evolution of DT systems
Problems
Research Goal
How to reuse existing engineering models
for developing DTs?
How to connect DT services to both
PT and simulations in a uniform way?
How to manage the creation and
evolution of DT systems?
Design
Questions
10

11. Architecture of the DT++ Framework
11
DT++ framework
Physical Twin
Engineering Model
Engineering Modeling
Language
«conformsTo»
Legend
Component
Artefact
Interface
Provided
Interface
Required
Interface
Inheritance
Component/Artefact
introduced by the
DT++ framework
«represents»

12. Contribution 1
12
Contribution 1:
Transformations between
Engineering and DT Models
DT++ framework
Physical Twin
Engineering Model
DT Modeling
Language (DTL)
Engineering Modeling
Language
M2M Transformation
«uses»
«uses»
«conformsTo»
DT ModelTransformation Engine
«input» «output»
«uses»
Legend
Component
Artefact
Interface
Provided
Interface
Required
Interface
Inheritance
Component/Artefact
introduced by the
DT++ framework
«conformsTo»
«represents»
[1] Lehner, D., Sabine, S., Vierhauser, M., Narzt, W., & Wimmer, M. (2021). AML4DT: A model-driven framework for developing and maintaining digital twins with
AutomationML. International Conference on Emerging Technologies and Factory Automation (ETFA).
[2] Lehner, D., Eisenberg, M., Sint, S., & Wimmer, M. (2023). A pattern catalog for augmenting Digital Twin models with behavior.at-Automatisierungstechnik,71(6).

13. Contribution 1: AML2DTDL Transformation
13
DTL
CAEX
Language
AutomationML
Model
DT Model DTDL Model
Transformation
Engine
Transformation
Engine
DTL2DTDL
Mapping
Digital Twin Definition
Language (DTDL)
CAEX2DTL
Mapping
«conformsTo»
«conformsTo» «conformsTo»
«uses» «uses»
«uses» «uses» «uses» «uses»
«input» «input» «output» «output»
Lehner, D., Sint, S., Vierhauser, M., Narzt, W., & Wimmer, M. (2021, September). AML4DT: A model-driven framework for developing and
maintaining digital twins with AutomationML. In International Conference on Emerging Technologies and Factory Automation (ETFA).
Car
-manufacturer: string
-batteryLevel: int
-temperature: int
-toggleHeating()
-toggleCooling()
-toogleEnergyMode()
Interface „Car“ (DTMI: „dtdl.v2.com.carcompany.car:1“):
Property „manufacturer“ (String, isWritable=false)
Telemetry „batteryLevel“ (float)
Telemetry „temperature“ (int, type=„Temperature“,
unit=„Celsius“)
Command „toggleHeating“
Command „toggleCooling“
Command „toggleEnergyMode“
«is_a» «is_a»

14. Contribution 1: UML2DTDL Transformation
14
DTLUML
UML Model DT Model
Transformation
Engine
UML2DTL Mapping +
Language Embeddings
«conformsTo» «uses»
«uses»
«output»
«uses»
«input»
«conformsTo»
«input»
Lehner, D., Sint, S., Eisenberg, M., & Wimmer, M. (2023). A pattern catalog for augmenting Digital Twin models with
behavior.at-Automatisierungstechnik,71(6), 423-443.
Car
-manufacturer: string
-batteryLevel: int
-temperature: int
-toggleHeating()
-toggleCooling()
-toogleEnergyMode()
«StateMachine»
«State»
HeatingOn
«State»
HeatingOff
toggle
Heating()
toggle
Heating()
«is_a» «is_a»
Car CarStateMachine
HeatingOn HeatingOff
StateState
StateMachineStateMachine
«is_view»
StateState
StateMachineStateMachine

15. Contribution 1: Evaluation
Method: Case Study

RQ1.1 (Feasibility for structural model): Is it possible to represent a DT in a DT platform involving
reasonable adaptation effort to an initial AML model?
•Result: Possible with small adaptations
RQ1.2 (Effort): What is the effort of creating and maintaining a DT in a DT platform using the
DT++ framework, compared to a traditional setup with the DT++ framework?
•Result: Effort reduced by ~50 %
RQ 1.3 (Size of DT model with behavioral information): What is the DT model size for DTs with
encoded behavioral information for different model transformation patterns
•Result: Manageable sizes for all patterns
15
[1] Lehner, D., Sabine, S., Vierhauser, M., Narzt, W., & Wimmer, M. (2021). AML4DT: A model-driven framework for developing and maintaining digital twins with
AutomationML. International Conference on Emerging Technologies and Factory Automation (ETFA).
[2] Lehner, D., Eisenberg, M., Sint, S., & Wimmer, M. (2023). A pattern catalog for augmenting Digital Twin models with behavior.at-Automatisierungstechnik,71(6).

16. Methodological Overview
Reduce effort for engineering DT systems by employing MDE techniques
Redundant work when specifying DT models
in addition to engineering models
Redundant work when connecting DT
services to both PT and simulations
Lack of dedicated support for
managing the evolution of DT systems
Problems
Research Goal
How to reuse existing engineering models
for developing DTs?
How to connect DT services to both
PT and simulations in a uniform way?
How to manage the creation and
evolution of DT systems?
Design
Questions
•Adaptation effort for creating DT model
from AML model?
•Effort of setup and maintenance of a DT
in a DT platform?
•DT model size for DT with encoded
behavioral information?
Knowledge
Questions
Transformations between Engineering and
DT Models
Contributions
16

17. Contribution 2: Common
interface to connect services
to both DT and simulations
Contribution 2
17
Contribution 1:
Transformations between
Engineering and DT Models
DT System
DT++ framework
Physical Twin
Engineering Model
Twin Manager
DT Modeling
Language (DTL)
Engineering Modeling
Language
M2M Transformation
Simulation
DT Platform
«uses»
«uses»
interprets
«conformsTo»
DT ModelTransformation Engine
«input» «output»
communicates
«uses»
<<abstract>>
DT System Component
Legend
Component
Artefact
Interface
Provided
Interface
Required
Interface
Inheritance
DT Services
Component/Artefact
introduced by the
DT++ framework
«conformsTo»
«represents»
Lehner, D., Gil, S., Mikkelsen, P. H., Larsen, P. G., & Wimmer, M. (2023). An Architectural Extension for Digital Twin Platforms to
Leverage Behavioral Models. International Conference on Automation Science and Engineering (CASE).

18. Contribution 2: Twin Manager Approach
18
Twin
Manager
SimulationDT Platform
DT Model
DT Services
Azure
DT Platform
«wraps
«configures»
DTDL Model
Henshin
Engine
Henshin Model
AzureDTImpl HenshinImpl
«Interface»
Endpoint
Transformation
Engine
Transformation
Engine
«output»
«input»
«input»
«output»
«configures»
«wraps»
DTDL2DTL
Mapping
Henshin2DTL
Mapping
Event-based Monitor
«uses»
«configures»
Event-based Planner
«uses»
«communicates»
«communicates»
Lehner, D., Gil, S., Mikkelsen, P. H., Larsen, P. G., & Wimmer, M. (2023). An Architectural Extension for Digital Twin Platforms to
Leverage Behavioral Models. International Conference on Automation Science and Engineering (CASE).

19. Contribution 2: Evaluation
Method: Case Study
•DT++ framework VS DT platforms VS no platform support
RQ2.1 (Implementation Effort): What is the effort for implementing a simulation-based DT system?
•DT++ outperforms existing solutions in all scenarios
•Effort reduced by at least 56 %
RQ2.2 (Adaptation Effort): What is the effort for changing the DT services from one case study to
another?
•DT++ outperforms existing solutions in all scenarios
•Effort reduced by at least 39 %
19Lehner, D., Gil, S., Mikkelsen, P. H., Larsen, P. G., & Wimmer, M. (2023). An Architectural Extension for Digital Twin Platforms to
Leverage Behavioral Models. International Conference on Automation Science and Engineering (CASE).

20. Methodological Overview
Reduce effort for engineering DT systems by employing MDE techniques
Redundant work when specifying DT models
in addition to engineering models
Redundant work when connecting DT
services to both PT and simulations
Lack of dedicated support for
managing the evolution of DT systems
Problems
Research Goal
How to reuse existing engineering models
for developing DTs?
How to connect DT services to both
PT and simulations in a uniform way?
How to manage the creation and
evolution of DT systems?
Design
Questions
•Adaptation effort for creating DT model
from AML model?
•Effort of setup and maintenance of a DT
in a DT platform?
•DT model size for DT with encoded
behavioral information?
Knowledge
Questions
•Implementation effort for simulation-
based system?
•Adaptation effort for changing
underlying PT of a DT service?
Transformations between Engineering and
DT Models
Common interface to connect
services to both DT and simulations
Contributions
20

21. Contribution 3:
Method to manage
DT architecture
variability
Contribution 3
21
Contribution 1:
Transformations between
Engineering and DT Models
DT System
DT++ framework
Physical Twin
Engineering Model
Twin Manager
DT Modeling
Language (DTL)
Engineering Modeling
Language
M2M Transformation
Simulation
DT Platform
«uses» «uses»
interprets
«conformsTo»
DT ModelTransformation Engine
«input» «output»
Reference Architecture
Language
DT System
Generation Engine
Reference Architecture
Model
DT System Component
Language
DT System Component
Model
communicates
«uses»
«abstract»
DT System Component
«represents»
Legend
Component
Artefact
Interface
Provided
Interface
Required
Interface
Inheritance
DT Services
Component/Artefact
introduced by the
DT++ framework
«conformsTo»
«represents»
«represents»
«uses»
«conformsTo» «conformsTo»
«input»
«output»
Contribution 2: Common
interface to connect services
to both DT and simulations
Lehner, D., Pfeiffer, J., Klikovits, S., Wortmann, A., & Wimmer, M (2024). A Method for Template-based Architecture Modeling and
its Application to Digital Twins. Journal of Object Technologies, 23(3).

22. Contribution 3: Details
22
Phase 3Phase 2Phase 1
Result: DT System
DT++ framework
Reference Architecture
Language
DT Generation
Engine
Reference Architecture
Model
DT System Component
Language
DT System Component
Model
«abstract»
DT System Component
Legend
Component
Artefact
Dependency
Phase of the proposed
method
«output»
«input»
«conformsTo»
«represents»
«represents»
«uses»
«conformsTo»
Lehner, D., Pfeiffer, J., Klikovits, S., Wortmann, A., & Wimmer, M (2024). A Method for Template-based Architecture Modeling and
its Application to Digital Twins. Journal of Object Technologies, 23(3).
Simulation DT Platform DT Services …

23. Contribution 3: Evaluation
Method: Case Study
•DT++ framework VS module-based approach VS MAPE-based approach
RQ3.1 (Expressiveness): What is the expressiveness of specifying DT systems using the proposed
template-based approach, compared to the state of the art?
•DT++ framework + module-based approach support all DT systems in our example
•MAPE-basedapproach supports only a subset (12/30)
RQ3.2 (Effort): What is the effort for defining and evolving DT systems using the DT++ framework
compared to the state of the art?
•DT++ framework only requires changeson the DT reference architecture level
•Module-basedapproach requires manual integration and adaptation of all DT systems
•MAPE-basedapproach is optimized for MAPE-based DT systems
23Lehner, D., Pfeiffer, J., Klikovits, S., Wortmann, A., & Wimmer, M (2024). A Method for Template-based Architecture Modeling and
its Application to Digital Twins. Journal of Object Technologies, 23(3).

24. Methodological Overview
Reduce effort for engineering DT systems by employing MDE techniques
Redundant work when specifying DT models
in addition to engineering models
Redundant work when connecting DT
services to both PT and simulations
Lack of dedicated support for
managing the evolution of DT systems
Problems
Research Goal
How to reuse existing engineering models
for developing DTs?
How to connect DT services to both
PT and simulations in a uniform way?
How to manage the creation and
evolution of DT systems?
Design
Questions
•Adaptation effort for creating DT model
from AML model?
•Effort of setup and maintenance of a DT
in a DT platform?
•DT model size for DT with encoded
behavioral information?
Knowledge
Questions
•Implementation effort for simulation-
based system?
•Adaptation effort for changing
underlying PT of a DT service?
•Expressiveness of Template-based
approach
•Effort for defining and evolving DT
systems
Transformations between Engineering and
DT Models
Common interface to connect
services to both DT and simulations
Method for template-based DT system
modeling
Contributions
24

25. Discussion
•DT++ Framework evaluated in multiple case studies
•Air quality exemplar
•Self-adaptive incubator
•Stack balancing system
•DT services for self-adaptive cars
+Effort for setup and maintenance of DT (RQ1) and setup and maintenance of DT services (RQ2)
reduced significantly compared to state of the art
+Provides both flexibility and scalability for managing multiple DT systems (RQ3)
−Drawback: Increased overhead for small number of DT systems
−Drawback: Increased overhead for DT systems with common reference architecture
25

26. Conclusion and Future Work
Successfully applied MDE to develop + maintain DT systems
•DT++ framework reduces effort by solving open problems in DT platforms
•Open-source repositories to facilitate adoption
Future Work
•Multivocal review of DT platforms
•Extend generalizability
•Contribution 1: Implement further transformations
•Evaluation: Structured experiments/action research
•Extend functionality of TwinManager
•Build DT template and module library
•Transfer learnings to existing DT modeling languages
26

27. Publications
Problem Investigation
•Govindasamy, H. S., Jayaraman, R., Taspinar, B., Lehner, D., & Wimmer, M. (2021). Air quality management: An exemplar for model-
driven digital twin engineering. International Conference on Model Driven Engineering Languages and Systems Companion
(MODELS-C), pp. 229-232, IEEE.
•Lehner, D., Pfeiffer, J., Tinsel, E. F., Strljic, M. M., Sint, S., Vierhauser, M., ... & Wimmer, M. (2021). Digital twin platforms: requirements,
capabilities, and future prospects. IEEE Software, 39(2), 53-61.
•Pfeiffer, J., Lehner, D., Wortmann, A., & Wimmer, M. (2022). Modeling capabilities of digital twin platforms-old wine in new bottles?.
Journal of Object Technologies, 21(3).
Systematic Mapping Study
•Lehner, D., Zhang, J., Pfeifer, J., Splettstößer, A., Sint, S., Wortmann, A., & Wimmer, M. (2025) Model-Driven Engineering for Digital
Twins: A Systematic Mapping Study. Journal of Software and Systems Modeling (accepted for publication) 2024
Solution + Validation
•Lehner, D., Sint, S., Vierhauser, M., Narzt, W., & Wimmer, M. (2021). AML4DT: A model-driven framework for developing and
maintaining digital twins with AutomationML. International Conference on Emerging Technologies and Factory Automation
(ETFA), pp. 1-8, IEEE.
•Lehner, D., Sint, S., Eisenberg, M., & Wimmer, M. (2023). A pattern catalog for augmenting Digital Twin models with behavior.at-
Automatisierungstechnik,71(6), 423-443.
•Lehner, D., Gil, S., Mikkelsen, P. H., Larsen, P. G., & Wimmer, M. (2023). An Architectural Extension for Digital Twin Platforms to
Leverage Behavioral Models. International Conference on Automation Science and Engineering (CASE), pp. 1-8, IEEE.
•Lehner, D., Pfeiffer, J., Klikovits, S., Wortmann, A., & Wimmer, M (2024). A Method for Template-based Architecture Modeling andits
Application to Digital Twins. Journal of Object Technologies, 23(3).
27

28. https: //cdl-mint. se. jk u. a t/
CDL-MINT
https://cdl-mint.se.jku.at/
Christian Doppler Laboratory for Model-Integrated Smart Production
CDL-MINT
Thank You!
Comments? Questions? Feedback?
Daniel Lehner
[email protected]
http://github.com/derlehner