Recursive virtual sensing methods in structural dynamics

Abstract

The ability to acquire precise and reliable information from objects and their surrounding environments has become increasingly critical in today's data-centric world. However, direct measurement of these quantities is not always feasible due to physical and financial constraints. Virtual sensing offers the possibility to estimate or predict quantities that are difficult, expensive, or impossible to measure directly. This cutting-edge technique serves as the fundamental tool for applications in structural dynamics, such as structural health monitoring, active vibration control, model updating and online digital twinning.

This work addresses challenges in virtual sensing by developing a robust virtual sensing framework. The proposed framework contains a series of model-based methods with universal applicability and different capabilities, enabling real-time system identification of partially observed systems across various sensor networks and under different levels of uncertainties. Numerical and experimental studies demonstrate significant improvements in performance compared to other state-of-the-art methods available in the literature.