PSI - Issue 84

Francesco Mariani et al. / Procedia Structural Integrity 84 (2026) 773–780

777

with seasonal variations. The relative humidity measurements (H1) show significant variability, with values generally ranging between approximately 50% and 95%, and higher fluctuations during the winter period. Some data gaps and isolated outliers can be observed, likely due to temporary sensor malfunction or data acquisition interruptions. Table 2 summarizes the investigation phases for the different structural configurations of the bridge during the monitoring program. The configurations differ in terms of the presence and layout of safety barriers and the prestressing condition of the post-tensioning cables. For Phases 1, 2, and 3, only Ambient Vibration Tests (AVT) were performed, whereas continuous monitoring was activated after October 24th for Configurations 4 and 5.

Table 2. Phases of the experimental investigation.

Configuration no.

Configuration description

Type of investigation

Dates

1

Bridge without safety barriers and without post-tensioning cables, not prestressed Bridge equipped with lateral safety barriers in contact with the deck but not rigidly connected, with post-tensioning cables not prestressed Bridge with safety barriers installed on only one side, with post-tensioning cables not prestressed Bridge with safety barriers installed on only one side, with post-tensioning cables prestressed Bridge without safety barriers, with post-tensioning cables prestressed

AVT

28 August 2025

2

AVT

17 October 2025

3

AVT

23 October 2025

4

Continuous monitoring Continuous monitoring

24 October – 6 November 2025

5

From 6 November 2025

3. Structural identification results The modal identification for the different structural configurations was performed using the Mova/Moss software suite (García-Macías and Ubertini, 2020). Among the implemented structural identification approaches, the SSI_Cov method has been used (Cancelli et al. 2020). This technique, known as Stochastic Subspace Identification, based on covariance data, is an output-only modal analysis approach that estimates the dynamic properties of a structure—such as natural frequencies, damping ratios, and mode shapes—directly from measured response signals under ambient or operational excitation. By analyzing data from AVTs, the first five vibration modes were identified. Due to the low level of ambient vibration in the experimental site, upper modes were not identified in a stable manner. Figure 4 shows the results of the Operational Modal Analysis performed on the bridge in reference condition (configuration 5). The upper-left panel reports the average Power Spectral Density (PSD) functions of the measured responses, with the identified modal frequencies highlighted. The stabilization diagram and peak-picking results allow the identification of five dominant modes in the frequency range up to about 25 Hz. The upper-right panel shows the Modal Assurance Criterion (MAC) matrix, confirming a good separation and consistency among the identified mode shapes, with high values along the main diagonal and negligible cross-correlation. The lower panels illustrate the reconstructed mode shapes corresponding to the first five vibration modes, together with their natural frequencies. The results demonstrate a clear and stable modal identification, providing a reliable reference baseline for subsequent comparisons under different structural configurations and boundary conditions. 4. Continuous monitoring results During the continuous monitoring (configurations 4 and 5) each 30-minute acquisition was processed to obtain time histories of natural frequencies and mode shapes. Due to the low level of ambient excitation, only three modes were consistently identified in each acquisition and were therefore selected for tracking. To account for environmental variability and to detect outliers in the modal parameters, a statistical model based on Principal Component Analysis (PCA) was developed. PCA was applied to the three tracked natural frequencies, yielding two principal components that together explain approximately 99% of the total variance. These components provide a reduced representation of