The Digital Twin concept has been around since the early 2000’s. Where it was first adopted in the space and aircraft industry by actors like NASA and the U.S. Airforce, it gradually found its way into other application domains, like Industry 4.0, Smart Manufacturing, Healthcare or Smart Cities. NASA applied it in its technology roadmap on modelling, simulation, information technology and processing and defined it as “an integrated multi-physics, multi-scale probabilistic simulation of a vehicle or system that uses the best available physical models, sensor updates, fleet history, etc., to mirror the life of its flying twin“ (Shafto et al. 2012).
Since then, definitions have been altered and challenged by different authors. Boschert and Rosen (2016) perceived it as a description of a component, product or a system that evolves with the real system. Grieves and Vickers (2017) defined it as a set of virtual information constructs that fully describes a potential or actual physical manufactured product from the micro atomic level to the macro geometrical level. According to Alam and El Saddik (2017), the Digital Twin is “an exact cyber copy of a physical system that truly represents all of its functionalities”. Finally, Zheng et al. (2019) term the Digital Twin as “an integrated system that can simulate, monitor, calculate, regulate, and control the system status and process”.
Thus, different definitions of Digital Twins exist, but the simple definition we will be using here is to see Digital Twins as a “Virtual representation of a physical entity with a bi-direction communication link”. Indeed, according to Grieves & Vickers (2017), an essential part of a Digital Twin is that between the virtual and non-virtual entity, there exists a bi-directional data link. This bidirectionality entails that data flows from the physical to the virtual world to support sensing and from the virtual to the physical to support actuation. Indeed, a Digital Twin can be seen as being part of a cyber-physical or an internet-of-things system of which sensing and actuation are salient features. That there is a close data relationship between the physical and virtual entities allows the virtual Digital Twin to evolve along with its non-virtual counterpart, mirroring its lifecycle. This allows the Digital Twin to predict future states of the physical entity and allows simulating and testing novel configurations of the physical entity. A Digital Twin is thus not merely a detached virtual representation of a physical twin, but rather an evolving representation that interacts with its physical twin via sensors and actuators connected through Internet of Things technology.
Regarding recent Digital Twin applications, Brenner and Hummel (2017) developed a digital shop floor management system based on the Digital Twin notion. A three-dimensional Digital Twin was devised by Knapp et al. (2017) in manufacturing to predict variables affecting metallurgical structures. However, the digital representation was not connected to its physical counterpart via sensors, so the presented Digital Twin notion does not clearly correlate with the definitions presented earlier. Schleich et al. (2017) proposed a simple reference model for Digital Twins in product design and manufacturing. In the work of Söderberg et al. (2017), the Digital Twin is referred to as a simulation for real-time control and optimization of products and production systems, where the authors specify data models to move from mass to a more individualized production. Alam and El Saddik (2017) developed a driver assistance application where the Digital Twin is used to identify various driving events and provide recommendations for drivers, insurance companies and emergency units. Uhlemann et al. (2017) introduced a learning factory based on a Digital Twin to demonstrate its benefits and familiarize the workers with new technologies and their implementation. Tao et al. (2018) focussed on how to generate and converge cyber-physical data and applied their framework in three cases that relate to product design, product manufacturing and product service. Lastly, Zheng et al. (2019) applied the Digital Twin concept to model a welding production line.
In terms of functional aspects of Digital Twins, Barricelli et al (2019) discussed the fact that an added value of Digital Twins lies in the combination of data sources that otherwise remain siloed. Indeed, doing so allows the creation of new decision-making insights by combining data from disparate domains, like the impact of changes to motorized ground traffic situations on air quality. Both air quality and traffic data exist in different databases and are administered by people belonging to different disciplines. Combing them allows for new types of insights to surface. Such combination of data sources requires the sources to be semantically annotated, which explains why ontologies are essential and constitute an active research field in Digital Twins (Wright and Davidson 2020). The same is true for data standardization, allowing data streams to be ingested by different Digital Twins (Datta 2016). In the built environment domain, data standards with semantics support like NGSI-LD1, Linked BIM (Bonduel, 2018), CityGML 3.02 are becoming key to bridging the semantic gap that exists between different data realms.
The last 5 years have seen great interest in the Digital Twin paradigm in many different areas, as well as in the logistic sector. The recent surge of interest in Digital Twins is due to the maturity of some of the essential technologies that enable it. Firstly, the link between the virtual and physical entities has been made increasingly affordable, ubiquitously deployable, stable and high-bandwith due to the development of new wireless communication protocols like 4G, 5G, Sigfox, LoRa and NB-IoT. Secondly, AI technologies such as supervised learning for the extraction of meaningful data from sensor-based data streams (timeseries, camera, audio, location,… ) and reinforcement learning for the evolution of effective behavioural patterns, have made great advances.
Open Urban Digital Twins
A recent addition to the application domain of Digital Twins is the Open Digital Twin architecture (Coenen, 2020). In this architecture,shown in the figure below, there are two main parts. The top part is the Digital Twin platform, which is used for supporting decisions. To do so, it has a Graphical User Interface (GUI), which allows users to read and write data to a semantic broker that forms the heart of the Digital Twin architecture and combines data coming from various sources. These sources are to be found in the bottom part of that architecture and are called Solution Accelerators, to be understood as a combination of data sources (IoT sensor stream or static data sources) that are treated by one or more algorithms (AI-based or more classical) and that are exposed through an API (preferably REST or GraphQL based). The data that is both produced and ingested by the Solution Accelerators needs to be standardized and semantically annotated (by applying ontologies) as much as possible. Doing so allows different parts of the architecture to communicate. Solution Accelerators exist in different domains and can be relevant to each other. For example, in the domain of Air Quality modelling, predictions on particulate matter pollution can be made based on predictions in the domain of traffic modelling, as knowing how many trucks, cars or motorbikes will be at a certain location allows predicting the amount of particulate matter pollution. Standardization and semantic annotation of Solution Accelerator outputs allows a particular motorized traffic prediction Solution Accelerator to be replaced by another one, without hindering the operations of the air quality prediction Solution
Accelerator that relies on it. This is why this architecture is termed “Open”: as long as the standards and semantics respect the general outlines imposed by the architecture, different parties can plug-in their own Solution Accelerators. Also, as Solution Accelerators are decoupled from the Digital Twin platform and hosted using e.g. their own Docker microcontainers, they can interact with other Digital Twin platforms.
Digital Twins at imec
Digital Twins have gained a lot of attention at imec in the past three years. First there is the focus on Urban Digital Twins, which has become one of the main focal points in the City of Things smart city program. There, Digital Twins are seen as a main platform for supporting decisions that pertain to the Urban context. Different cities (e.g. Bruges, Antwerp, Ghent) in Flanders are working with imec on Digital Twin projects. In addition, imec is involved in initiatives at the EU level to deploy Urban Digital Twins in various cities, like Pilsen, Athens, Helsinki and Santander.
Besides Urban Digital Twins, other imec teams are working on Digital Twins as well. At OnePlanet, research is underway to build Digital Twins of the human body. IDLab Antwerp is working on Digital Twins that can be used to better control aspects of buildings, like HVAC systems. Finally, EnergyVille teams are working on modelling of the yield of Photovoltaic installations. Imec’s interest in the Digital Twin concept stems from its ability to integrate various sensors in a way that allows end-users to make better decisions in diverse domains. It can integrate competences from many different imec teams, like sensor development, IoT communications, algorithm development, semantic modelling, GUI design and governance support. As it is a natural fit with its DNA, imec has decided to become member of the Digital Twin Consortium, which coordinates standardisation and valorisation of Digital Twin activities across domains (infrastructure, aerospace, health, …).
Data specifications and semantics for Urban Digital Twins
The aim of the Urban Digital Twin architecture, presented above, is to allow various algorithms, in various domains, to read and write data to and from the Digital Twin platform. In order to do this, data must be made understandable by a number of systems that may not be acquainted with each other’s sematic scope. Various technologies have emerged to tackle different aspects of this issue. Digital Twins build on real-time datastreams that are produced by local IoT stacks. In this domain, a lot of traction has been recently created in the area of datastream contextualisation, with e.g. the ETSI NGSI-LD3, W3C SSN/SOSA and W3C Web of Things specifications. In the world of buildings, which represent an essential part of cities, Building Information Modelling (BIM) standards and ontologies have been developed and deployed at a rapid pace. For example, Bentley Systems’ iModel.js4 provides an open way to defined Digital Twins of buildings.
For urban modelling, standards like CityGml have been popular for a while and are being continuously improved. CityGML 3.05 promises to offer more semantic possibilities, as well as the embedding of datastreams at specific geospatial locations. A very active player in the domain of Digital Twins technology is Microsoft, who recently came out with its v2.0 version of the Digital Twin Definition Language (DTDL). Finally, the Onnx standard offer interesting opportunities for the interoperability between machine learning frameworks. Supervised learning and reinforcement learning are essential techniques for the creation of Solution Accelerators and are set to become even more important in the coming years.
In Flanders, the OSLO (Open Standard for Linking Organizations) initiative is creating Linked Data vocabularies for different domain models, as well as application profiles, which define how to use a vocabulary for a certain registry (Buyle, 2019). The semantic interoperability which is created this way across Flemish datasets creates a decentralized knowledge graph across the Flemish government. Datasets such as the road registry, points of interest, waterways, Flemish “subsoil” database, etc. have become semantically annotated, and can thus be programmed to be automatically adopted and integrated in Urban Digital Twins.
In order to describe algorithms and functions, the Function Ontology6 was developed in prior work (De Meester, 2016). It introduces a data model, specification, and ontology to semantically declare and describe functions independently of the used technology. This way, it allows to declare and use actionable events in semantic applications, without restricting the use to programming languagedependent implementations. The ontology allows for extensions, and is proposed as a possible solution for semantic applications in various domains.
Urban Digital Twins are applicable to various domains (air quality, traffic, people flows, water quality, flood prediction, …) which each have their own logic and semantic scope. The main question is: “How can developers of Urban Digital Twins make their algorithms interoperable across different domains and IoT datasets?”. For example, how can an air quality model use the data that is produced by a traffic model, to make predictions on air quality, based on a prediction of the amount of motorized vehicles at a specific location?
In order to allow such interoperability between the various algorithms and the Urban Digital Twin platform, the PhD student will study four entangled sub-problems:
(i) Describing predictive models
This has been described as one of the big opportunities in Digital Twin design by the Open Data Institute (O’Donnell, 2019 ). However, no concrete solution has been formulated yet. The PhD student will research whether the Function Ontology (De Meester, 2016) can be extended to describe Digital twin Algorithms.
(ii) Data publishing strategies for timeseries
Once semantically described using e.g., NGSI-LD or W3C SSN/SOSA, publishing timeseries in an API still poses multiple challenges such as, how much of the history needs to be retained and published, what statistical summaries must be calculated by the server, or how to choose an optimal sampling strategy. Predictive models can rely on the timeseries’ properties to speed up querying, and reliably compare timeseries against each other. One such technique is the recent Matrix Profile (Imani 2018) that allows to precompute discords, motifs and snippets for enabling faster processing of data for timeseries analysis. The PhD student will extend the state of the art and apply these techniques within a Linked Data context.
(iii) Describing derived datasets
When a Digital Twin is confronted with multiple datasets, a query planner needs to be able to automatically select the right datasets for the model. Therefore, it needs to understand whether
certain timeseries are derived from each other and in what way. The PhD student will study how to describe derived timeseries and how to publish these in a Web API.
(iv) Multi-layered Digital Twin architectures
Combining the ideas described in (I), (ii) and (iii), the PhD student will be able to devise a multilayered digital twin architecture, where all involved stakeholders can play their part.
In prior work, Imec experimented in the context of Microsoft’s Azure Digital Twin infrastructure as well as in the Digital Twin Consortium’s7 Taxonomy and Semantics workgroup. Further expanding this line of inquiry could allow imec to acquire a central position in the Urban Digital Twin domain, both locally in Flanders, where Urban Digital Twins are rapidly gaining traction, as well as internationally.
- Alam, K. M., & El Saddik, A. (2017). C2PS: A digital twin architecture reference model for the cloud-based cyber-physical systems. IEEE access 5, 2050–2062
- Barricelli, B. R., Casiraghi, E., & Fogli, D. (2019). A Survey on Digital Twin: Definitions, Characteristics, Applications, and Design Implications. IEEE Access, Vol 7, 167653-167671.
- Bonduel, M., Rasmussen, M. H., Pauwels, P., Vergauwen, M., & Klein, R. (2018). A novel workflow to combine BIM and Linked Data for existing buildings. In eWork and eBusiness in Architecture, Engineering and Construction: Proceedings of the 11th European Conference on Product and Process Modelling (ECPPM 2018). CRC Press.
- Boschert S., & Rosen, R. (2016) Digital Twin—The Simulation Aspect. In: Hehenberger P., Bradley D. (eds) Mechatronic Futures (pp. 59-74). Springer, Cham.
- Brenner, B., & Hummel, V. (2017). Digital twin as enabler for an innovative digital shopfloor management system in the ESB Logistics Learning Factory at Reutlingen-University Procedia Manufacturing 9, 198– 205.
- Buyle, R., Vanlishout, Z., Coetzee, S., De Paepe, D., Van Compernolle, M., Thijs, G., Van Nuffelen, B., De Vocht, L., Mechant, P., De Vidts, B., others: Raising interoperability among base registries: The evolution of the Linked Base Registry for addresses in Flanders. Journal of Web Semantics. 55, 86–101 (2019).
- Coenen, T. (2020), Built Environment Digital Twin, lecture at Cambridge University Center for The Digital Built Britain [Video]. Youtube. https://www.youtube.com/watch?v=HEY2c0Nik0Y
- De Meester, B., Dimou, A., Verborgh, R., & Mannens, E. (2016, May). An ontology to semantically declare and describe functions. In European Semantic Web Conference (pp. 46-49). Springer, Cham.
- Grieves, M., & Vickers, J. (2017). Digital twin: Mitigating unpredictable, undesirable emergent behavior in complex systems. Transdisciplinary perspectives on complex systems, 85–113. Springer.
- Imani, S., Madrid, F., Ding, W., Crouter, S., & Keogh, E. (2018). Matrix Profile XIII: Time Series Snippets: A New Primitive for Time Series Data Mining. In 2018 IEEE International Conference on Big Knowledge (ICBK) (pp. 382-389). IEEE.
- Knapp, G., T. Mukherjee, J. Zuback, H. Wei, T. Palmer, A. De, and T. DebRoy (2017). Building blocks for a digital twin of additive manufacturing. Acta Materialia 135:390–399.
- O’Donnell, F. (2019), Sharing models between ‘digital twins’, https://theodi.org/article/sharing-models-between-digital-twins/
- Schleich, B., N. Anwer, L. Mathieu, and S. Wartzack. 2017. Shaping the digital twin for design and production engineering. CIRP Annals 66(1):141–144.
- Shafto, M., M. Conroy, R. Doyle, E. Glaessgen, C. Kemp, J. LeMoigne, and L. Wang. 2012. Modeling, simulation, information technology & processing roadmap. National Aeronautics and Space Administration.
- Söderberg, R., K. Warmefjord, J. S. Carlson, and L. Lindkvist. (2017). Toward a Digital Twin for real-time geometry assurance in individualized production. CIRP Annals 66(1):137–140.
- Uhlemann, T. H.-J., C. Schock, C. Lehmann, S. Freiberger, and R. Steinhilper. 2017. The digital twin: Demonstrating the potential of real time data acquisition in production systems. Procedia Manufacturing 9:113–120.
- Wright, L., Davidson, S. (2020). How to tell the difference between a model and a digital twin. Advanced Modeling and Simulation in Engineering Sciences, 7(1), 1-13.
- Zheng, Y., S. Yang, and H. Cheng. 2019. An application framework of digital twin and its case study. Journal of Ambient Intelligence and Humanized Computing, 10(3), 1141-1153