PSI - Issue 62

S. Anastasia et al. / Procedia Structural Integrity 62 (2024) 1061–1068 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

1065

5

(a)

S1

S2

S3

S4

S5

S6

(b)

S1

S2

S3

S4

S5

S6

21.00m

21.48m

21.00m

21.12m

21.48m

21.12m

Fig. 1. (a) View of the Santa Ana viaduct, (b) arrangement of the SHM system's accelerometric sensors installed in the building.

3.2. Structural health monitoring (SHM) system The Structural Health Monitoring (SHM) system integrated into the structure excels not only in capturing vibrations induced by passing trains but also in recording deformations, displacements, wind, and temperature variations. For the assessment of vibrational behavior, uniaxial accelerometers were employed. Specifically, seventeen B&K 4507-B-006 uniaxial accelerometers, with a sensitivity of 490 mV/g, were strategically positioned in spans S1, S2, S5, and S6 (isostatic), as well as spans S3 and S4 (hyperstatic). This sensitivity level ensured avoidance of saturation during train passages. The accelerometers were oriented in the vertical direction and strategically placed at specified points Fig. 1b. Notably, they were deliberately positioned off the longitudinal axis of the viaduct, located on one of the side trusses. This placement aimed to associate obtained frequencies with both vertical and torsional vibrational modes. Automated measurements, synchronized with train passages, were facilitated by two photocells strategically installed before and after the bridge. Each measurement duration was 30 seconds, with a sampling frequency of 200 Hz, ensuring comprehensive data collection during these brief intervals. This approach leveraged the trains themselves as the excitation source for the structure, significantly reducing the time required for modal frequency acquisition compared to methods relying on ambient excitation. To ensure an accurate analysis of the structure's vibrational behavior, a crucial step was taken to separate the components related to the bridge's response from those attributed to the train, focusing solely on the tail generated by the excitation. Specifically, attention was directed to the last 10 seconds of recorded data, representing the tail of the train. It's important to note that this study does not encompass the entire monitoring period but rather focuses on a dataset corresponding to a single monitoring year. In particular, the analysis spans 7 months, from September 28, 2021, to May 2022, excluding several inactive periods observed in May 2021, September, and December 2022 during data collection.The estimation of natural frequencies employed the proposed automated procedure for data-driven Subspace System Identification (SSI) and extended for Covariance-driven Subspace System Identification (COV-SSI) as detailed in Section 2.1. Modal identification involved exploring model orders ranging from 150 to 250 in increments of 2. To discern structural poles from spurious ones, a hierarchical clustering approach, as outlined by (Enrique García Macías, 2022) was applied. 3.3. Feature extraction