Konstantinos Vlachas

Research area

Parametric reduced-order modeling for virtualization of nonlinear dynamical systems

Because of the emerging era of digital transformation, the notion of virtualization, or in other words, the establishment of a digital twin, has gathered increased attention. Here, the term digital twin refers to a virtual duplicate of an engineering system or mechanical component, relying on the concept of data assimilation in numerical representations of real-life systems. This coupling could, in turn, enable the prediction of the behavior of the system, thus offering insights regarding design optimization, operational decisions, or damage assessment.

When aiming to build a virtual representation of an actual real-life system, physics-based Reduced Order Models (ROM) offer a powerful idea. ROMs are low-order yet accurate representations of a high-fidelity system that can efficiently reproduce the dynamic physical response. My work focuses on developing a ROM framework to treat large-scale nonlinear dynamical systems or mechanical components, paving the road toward a digital twin representation. Such a scheme will treat parametric dependencies on structural traits and temporal and spectral characteristics of the applied excitation.

In this sense, we develop methodologies appropriate for model order reduction of nonlinear systems, which can also account for parametric dependencies and environmental and modeling uncertainties. Dynamic substructuring techniques can, in turn, be coupled to the ROM framework, thus enabling the treatment of intricate systems, where separate components may require individual reduction. Finally, we explore the potential of fusing monitoring data into the ROM framework to optimize its performance and achieve a robust approximation suitable for vibration monitoring and damage estimation applications.