Models Meet Data: Challenges to Create Virtual Entities for Digital Twins

Abstract

In recent years, digital twin (DT) technology has moved to the center of attention of many researchers and engineers. Commonly, a digital twin is defined based on a virtual entity (VE) that exhibits similar behavior to its physical counterpart, and that is coupled to this physical entity (PE). The VE thus forms a core part of any digital twin. While VEs may differ vastly—from ones based on simple simulation to high-fidelity virtual mirroring of the corresponding PE—they are typically composed of multiple models that may originate from multiple domains, address different aspects, and are expressed and processed using different tools and languages. Furthermore, the use of time series data—whether historical or real-time or both—from the PE distinguishes VEs from mere simulations. As a consequence of the modeling landscape complexity and the data aspect of VEs, the design of a digital twin and specifically of the VE as part of it represents several challenges. In this paper, we present our vision for the development, evolution, maintenance, and verification of such virtual entities for digital twins.