Do digital twins require physical simulations? A study of developing digital twins with varying reliance on physics-based models

Digital Twins (DTs) rely on the alignment of a physical asset and its digital representation to provide a useful insight into its operation. By bridging the gap between a physical entity and process data collected in real time, DTs have been at the forefront of the most recent wave of industrial digitisation. Due to the great variety of assets which can be modelled with a DT, there is no standardised way to develop one. Successful DT implementations found in literature vary from being strongly dependent on physical simulations, to being solely data-driven. Therefore, the consideration of the DT design process addressed in this work is necessary for incorporating physics-based models. To this end, we group DTs into three categories, namely: purely data-driven DTs, physics-based DTs and physics-informed DTs. We choose representative cases from literature to explain their distinguishing features, describe the intrinsic differences, as well as draw conclusions on their advantages and limitations. We then present the case study of developing a DT for FastBlade, a facility for regenerative testing of tidal turbine blades. We discuss the challenges and opportunities associated with the facility to assess the suitability of each of the three distinct DT development strategies. The complexity of the energy-recovery system, unknown asset internals, as well as the broad scope of the sensing and logging network are identified to be the key decisive factors. Finally, we suggest developing a physics-informed DT for FastBlade as the optimal route.