Digital Twin Development Framework in the context of Fluid Structure Interaction

Digital Twin (DT) is a relatively new concept of creating a virtual counterpart (digital object) to a physical system, initially introduced in the context of manufacturing by Grieves. Since its introduction the term "Digital Twin" has been used rather broadly, leading to the distinction by Kritzinger et al. (2018) of Digital Model and Digital Shadow based on the degree of automation in data flow between the physical and digital objects. Meanwhile, Wagg et al. (2020) introduced the concept of the hierarchy of digital twins based on the operational capabilities. The common feature of various definitions is a continuous feedback loop between virtual and physical system, often with a scope of achieving a higher accuracy of predictive capabilities and more rapid and informed decision-making (Van der Valk et al., 2020). In the context of wind energy, digital twins are seen mainly as a tool for more effective asset
management such as lifetime prediction (Tao et al., 2018; Branlard et al., 2020). However, DT technology can offer much wider potential applications ranging from design phase to specific operational improvements such as noise reduction, rotor imbalance or yaw misalignment detection.
While individual existing technologies such as Finite Element Analysis (FEA), Computational Fluid Dynamics (CFD), Machine Learning (ML), Bayesian methods, Multi-fidelity approaches already enable the tasks required of a DT such as verification and validation, physics-based modeling, data-augmented modelling and uncertainty quantification (Peherstorfer et al.,2018; Xiao and Cinnella, 2019; Renganathan et al., 2020), one of the main open challenges is the
creation of a system that enables the interaction between the above-mentioned components (Wagg et al., 2020). Such a system is usually referred to as workflow.
Unlike the majority of current wind turbine DTs that rely on existing SCADA data and strain gauge sensors, the twin discussed in this work integrates the data from pressure and acoustic sensors significantly expanding the possible use cases. Practical realisation of Supervisory and Operational pre-Digital Twins is straightforward with current software tools, meanwhile the development of full DTs is particularly dependent on availability and usability of fluid-structure interaction (FSI) simulation software packages with other components of the DT such as measurement system, machine learning modules, etc.
In terms of IT infrastructure, several commercial packages such as ANSYS Workbench enable automation of FSI simulations and subsequent creation of DTs via ANSYS Twin Builder, while there is lack of well developed open-source alternative available to the researches. A possible solution is the integration of packages such as OpenFOAM,
OpenFAST, XFOIL via louse coupling into the workflow through so-called wrappers. These wrappers can be seen as individual twin modules. In this regard, Octue platform offers a clear framework that specifies what each module can and/or must contain, as well as functions to validate incoming data and outputs, in addition to functions to check that a module itself is valid. Moreover, this framework deals with the task of combining individual modules together in hierarchies and networks. Further research in academic setting clearly necessitates the development of open-source libraries specifically facilitating implementation of DTs that incorporate FSI simulations. Moreover, realisation of logical frameworks such as the ones presented by Abdallah et al. (2017) and Bi and Zhao (2004) will be crucial for the future development of Digital Twins.