Modeling Capabilities of Digital Twin Platforms – Old Wine in New Bottles?

Digital twins are seen as core technologies to tackle the growing complexity of cyber-physical systems to better understand, monitor, and optimize their behavior. Digital twin platforms aim to facilitate the systematic engineering of digital twins by providing dedicated languages and corresponding tools to describe their abilities. However, with the emergence of these languages for digital twins, the question arises what the nature of these languages is and how they differentiate from existing modeling languages already used in the area of cyber-physical systems. To shed more light on this new modeling area, we study in this paper the modeling capabilities of three industrial digital twin platforms and frame them in existing and well-known modeling concepts provided by UML. In particular, we (i) extract the conceptual metamodels of three industrial digital twin platforms, (ii) compare them with common object-oriented modeling concepts of UML, (iii) and provide first insight about the portability of models between the platforms by performing an experiment. In particular, we use UML class diagrams as an anchor for relating the modeling concepts of digital twin platforms and as pivot for DT platform portability. Our investigation summarizes current modeling capabilities of digital twin platforms to provide a better understanding of their shared concepts to developers using such platforms. It also shows that these modeling capabilities often rely on well-known modeling concepts, but also add some new aspects. The performed experiment additionally gives first insights into the portability of different DT platform metamodels. To sum up, this work can be see as a starting point for uncovering the nature of digital twin modeling and providing a digital twin language family enabling developers to select appropriate modeling features for describing different aspects of digital twins without having to reinventing the wheel.