by Edgar Weippl (SBA) and Benjamin Sanderse (CWI)
This special issue addresses the current state of digital twin research, illustrating the many facets of this growing scientific field. The contributions collected here give insight in ongoing projects and allow for a glimpse into the future of digital twin technology.
In engineering, the use of models to represent reality might be as old as the profession itself. Most famously, NASA has built so-called “twins” of its spacecrafts since the early Apollo program: in 1970 Mission Control engineers improvised a device (nicknamed “the mailbox”) which was replicated by the Apollo 13 crew in order to solve one of the technical problems in their return to earth. Recently, the explosion in terms of data, algorithms, and computational power has allowed the creation of a new “twin” concept: Digital Twins.
A digital twin (sometimes also “digital shadow”) is a digital replica of real-world devices, processes or even persons. The technology draws on domains like machine learning, artificial intelligence and software analytics to provide a dynamic digital representation of its physical counterpart. Thereby, it uses data provided for example by Internet of Things (IoT) sensors as well as information coming from past machine usage and human domain experts. Currently mainly used in the context of Industry 4.0, digital twins provide models and simulations of, e.g., wind turbines or aircraft engines. Large companies like GE, Siemens, Shell and SAP are using this technology to create virtual models to monitor and diagnose their physical assets and systems, optimise operation and maintenance and calculate future performance . Especially in the context of cyber-physical systems (CPS), digital twins are of increasing importance.
The general idea is to create a digital partner throughout the lifecycle of all entities involved; digital twins are created entirely based on the specifications of their physical counterpart, whereby they document all its changes and developments. In order to do so, digital twins require data obtained from a system’s or device’s history, the experts working in the domain and even data from other (third party) entities, processes, and systems. Thus, a digital twin is able to provide information about and current status reports of its physical counterpart. Once set up, digital twins are used in manifold ways: a tool to handle data (e.g. generated by IoT devices), a model to run calculations and scenarios to reduce time and costs of product development and/or installation of complex systems, or a tool to provide a constant overview of systems which are often spatially distributed and used by multiple parties. Lastly, given that digital twins may easily be run in isolated environments, they can be meticulously analysed (e.g., regarding security measures) without disrupting operational systems. In the future, they will be joined with further technologies like augmented reality or AI capabilities, facilitating looking inside the digital twin, holding the promise to make checking the actual devices or production processes unnecessary .
The progress in developing digital twins is, amongst others, made possible through many important ongoing scientific research achievements in several fields in the mathematics and computer science community, such as: data assimilation, model order reduction, data-driven modelling and machine learning, high-performance computing for real-time simulation, visualization, etc. The articles in this special issue show that indeed much progress has been made, but that at the same time new research and new algorithms are imperative to make real-time simulation, data handling, and optimization of complex systems possible.
The industrial drive behind many digital twins is apparent in several articles in this issue. For example, Ponsard et al. discuss the case of a steel factory and demonstrate how a digital twin may be used for decision making even if the connection with the physical counterpart should partially or temporarily fail. Verriet et al. work with major industry players such as Philips Lighting, to develop an open architecture solution for connected lighting systems. The digital twin representing this system includes the environment in which it will be operating as well as the interactions with other systems and is used for validation and testing before actual installation. Strohmeier et al. present a messaging system which is the foundation for digital twins of industrial assets; it collects, monitors and analyses life cycle data with the goal of improving maintenance operations and long-term asset strategies. Further industrial applications are found in the articles of van Kruijsdijk (Shell), and Boschert and Rosen (Siemens).
Next to a clear industrial interest there is an important societal component in digital twin research. For example, Závodsky et al. work on developing human digital twins in order to provide optimal, personalized medical treatment of patients.
In many articles, security and safety play a key role in digital twin development. Boschert and Rosen look at transport infrastructure, in particular railway switches whose maintenance is essential to guarantee safety. By using a combination of measurement data and physics-based simulations they develop a digital twin which enables to identify failures in the physical components before they become critical. Eckhart and Ekelhart use virtual replicas of cyber-physical systems (CPSs) to monitor, visualize and predict the behaviour of CPSs. Their goal is to demonstrate how digital twins can increase the security of CPSs throughout their entire lifecycle. The authors have developed an experimental prototype and explicitly distinguish in their approach “CPS Twinning” in simulation mode (i.e., the digital twin runs independent from their physical counterparts) from replication mode (i.e., the twin mirrors the state of the physical device). Tauber and Schmittner look at security and safety evaluation in Industry 4.0 use cases. Given that system properties like security and safety are difficult to measure (as opposed to physical features), the authors investigated the modelling of such dependencies in relation to transparency and self-adaptability. Here, digital twins are a means to organise and manage all the data generated by IoT, since other models currently used are too static to accurately represent the dynamic and changeable nature of IoT devices. The authors also introduce the aspect of using digital twins in case of legal issues.
Lastly, the topic of security and safety is addressed by Damjanovic-Behrendt, who turns her attention on the Smart Automotive sector where strategic alliances between manufacturers make it difficult for researchers to gain access. Therefore, the collaborators in the project IoT4CPS are developing an open source digital twin prototype using machine learning for behavioural analysis and to predict security, privacy and safety measures.
Overall, the articles of this issue paint a detailed picture how industrial applications and academic research of digital twins are currently evolving and diversifying. The novelty of the topic of digital twins makes that many research activities are ‘work-in-progress’. In the coming years, we expect a flourishing research field in which mathematicians and computer scientists are working together to address important societal and industrial challenges.
 Digital twins – rise of the digital twin in Industrial IoT and Industry 4.0
• Thomas H.-J. Uhlemann, Christian Lehmann, Rolf Steinhilper, The Digital Twin: Realizing the Cyber-Physical Production System for Industry 4.0 (2017), https://doi.org/10.1016/j.procir.2016.11.152
• Roland Rosen, Georg von Wichert, George Lo, Kurt D. Bettenhausen, About The Importance of Autonomy and Digital Twins for the Future of Manufacturing (2015), https://doi.org/10.1016/j.ifacol.2015.06.141
• Digital twin technology and simulation: benefits, usage and predictions 2018,
Edgar Weippl, SBA Research, Austria
CWI, The Netherlands