Structural health monitoring system based on digital twins and real-time data-driven methods

In the era of technological acceleration marked by cloud computing, wireless communication, and digital modeling, the paper explores the intersection of these advancements through the lens of digital twins. In addressing the imperative of safe operation and maintenance in engineering projects, the paper proposes a groundbreaking structural health monitoring (SHM) system. This system synergizes the high-fidelity behavior simulation capabilities of digital twins with the potent data mining capabilities of intelligent algorithms. Highfidelity behavior simulation refers to the real-time, detailed replication of the structural dynamic characteristics in the digital twin environment. Structured in three integral parts, the SHM system offers real-time modal parameter calculation, comprehensive bridge estimation, and a data-driven warning methodology. The first component involves real-time modal parameter calculation for bridge structures, utilizing Bayesian modal identification based on acceleration data collected during operation. Concurrently, the second component extends the estimation to cover the entire bridge by employing Bayesian statistical inference and sequential importance resampling within the Kalman filtering framework. This facilitates the online estimation of system noise parameters and the reconstruction of responses at unmonitored positions. Finally, the third component introduces a data-driven warning method based on principal component analysis, combining pure data-driven techniques with single-indicator warning methods. The proposed system is exemplified through a case study focusing on a city bridge. The system not only achieves real-time monitoring but also provides warnings throughout the structure's entire lifecycle, addressing the urgent need for safe operation and maintenance in engineering projects. The study contributes to the evolving field of SHM by leveraging digital twin technology and intelligent algorithms for a robust and efficient monitoring system.