Structural Health Monitoring of Civil Infrastructures Using Innovative Techniques

Infrastructure plays a pivotal role in community development, exerting an important influence on economic growth. Structural Health Monitoring (SHM) is a multidisciplinary field encompassing the monitoring of infrastructure condition and integrity. Within this field, Operational Modal Analysis (OMA) offers a wide range of vibration-based monitoring solutions to address concerns regarding the reliability and integrity of civil assets while ensuring uninterrupted serviceability. OMA employs procedures to extract damage-sensitive features, indicating the integrity of the assets, without any need to measure the input excitation. This thesis focuses on adapting Automated Frequency Domain Decomposition (AFDD) using the Modal Assurance Criterion (MAC) to acquire essential modal properties such as natural frequencies and mode shapes. In the first part, after formulating the adapted methodology, the optimal performance of AFDD is established through a comprehensive sensitivity analysis. The analysis considers various factors influencing the extracted modal properties to develop a robust procedure. These influential factors include noise levels, spatial resolution of sensors, recording duration, and variation of hyperparameters present in the methodology. In this regard, field measurements from a cable-stayed bridge are analyzed by AFDD, optimizing the method by constructing stabilization diagrams. In these diagrams, the extracted natural frequencies from AFDD are compared with the data corresponding to the Finite Element (FE) counterparts. Based on the accuracy of extracted properties, the optimal ranges for each hyperparameter present in the AFDD are determined. Furthermore, this thesis presents an extensive output-only modal identification of various infrastructures using three distinct case studies: the Yonghe cable-stayed bridge, PolyU footbridge, and Moletta tower in the Maximus archaeological site. The dynamic characterization is performed by traditional Frequency Domain Decomposition (FDD), optimized AFDD, and covariance-driven Stochastic Subspace Identification (cov-SSI). The accuracy of the optimized AFDD method is evaluated, along with the potential limitations of each approach. Specifically, the effectiveness of these methods in identifying closely spaced, weakly excited modes, dealing with spurious peaks, and accurately identifying complex ones is examined. The study reveals valuable insights for each case study and highlights the risks of failing to identify particular vibrational modes when implementing OMA procedures. In summary, this thesis significantly contributes to the field of OMA by presenting an optimized AFDD approach for the long-term extraction of modal properties while comparing its performance with other well-defined methods. The findings from the case studies shed light on the strengths and limitations of each applied approach, offering valuable insights for the health assessment and monitoring of civil infrastructure.