Representing Industrial Data Streams in Digital Twins using Semantic Labeling

Digital twins provide the opportunity to observe and simulate current and future behaviour of cyber-physical systems in a non-invasive way. At the same time, digital representations of real-world assets allow to compare multiple situations in parallel and to estimate effects of planned interventions. Such simulations can be used in cyper-physical systems to make predictions about the future and prevent failures. However, it is complex to build a model that accurately represents the real-world and connects real-world assets to technically-oriented data streams produced by these assets. In this paper, we present a novel approach to model socalled Industrial Data Streams. Such streams are produced by assets in the real world and are integrated on a messageoriented middleware. Streams are annotated using the concept of semantic labels, which allows to assign characteristics such as location or type of the underlying assets to data streams. Our contribution consists of a semantic schema representation that categorises similar data streams, allowing subscribers to consume similar data from multiple assets within a single data analytics pipeline. Therefore, our approach paves the way for a more intuitive management of digital twin representations from industrial assets.