PSI - Issue 78

Andrea Miano et al. / Procedia Structural Integrity 78 (2026) 1903–1910

1907

remaining 60% of the pixels, information of both ascending and descending orbits are available and the vertical and east-west components can be computed correctly. Actually, when more than one track for one (or both) the orbits is available, multiple approaches could be followed for using this information: all the possible combinations selecting one ascending and one descending orbit could be considered, or (based on a quality index) the most reliable ascending and the most reliable descending orbit can be selected. Finally it is possible to merge all information solving the system of equations with a Weighted Least Squares approach in order to weight more the more reliable tracks. For more details, readers can refer to Talledo and Saetta 2025.

Fig. 2a Overview of POC-a corridor (in blue) with indicated the bridges considered in the analysis (in red); b Example of merged InSAR data for a specific bridge near Mantova: average vertical velocity computed for the 2015-2021 time window. Finally, for each grid pixel a set of multi-criteria flags is activated: 1. Average vertical velocity higher than a threshold of 4 mm/yr; 2. Average horizontal velocity higher than a threshold of 4 mm/yr; 3. Ascending average velocity along the LOS higher than a threshold of 4 mm/yr; 4. Descending average velocity along the LOS higher than a threshold of 4 mm/yr; 5. Variation with respect to the previous time window larger than 100% for each of the mentioned data points. It is worth noting that the last criteria exploit the annual release of EGMS keeping track of significant differences in terms of average annual velocity. For this last criterium, in order to account for possible large percentage differences, a threshold of 0.5 mm/yr is considered below which the flag is not activated irrespective of the percentage difference. Finally, it is important to keep in mind that proposed thresholds should be considered as preliminary suggestions; specific values can be selected based on the bridge typology (e.g., based on slenderness) after more detailed analyses. As an example, Figure 3a, shows the flag according to the first criteria for a specific bridge located north of Mantova. It can be seen that a vertical velocity above the selected threshold is obtained in the bridge’s west portion.

Fig. 3a. Example of activated flags at pixel level for a bridge; b. POC-a corridor showing for each bridge if at least one flag is activated. As a final step, an upscaling procedure is performed bridge-by-bridge activating a flag if at least one of its grid pixels has an activated flag. According to such upscale, of the total dataset of 480 bridges, about 200 bridges show at least one flag for the last criteria (i.e., a significant percentage difference for two adjacent time windows), 65 bridges show the activation of a flag related to velocities along LOS higher that the threshold of 4 mm/yr, while 31 bridges show the activation of a flag related to reaching the threshold for vertical and/or east-west velocity components. Figure 3b shows a view of the whole POC-a corridor indicating in red the bridges with a flag activated. 2.2. POC 2-b: Roadway network in a mountainous region The parametric discretization of bridge footprint showed the strong influence of the assumptions for MPs clustering. Results from the two adopted procedures are compared in Figure 4a. The first approach (i.e., 30x30m cross section grid, without subdivision along the longitudinal axis of the bridge) allows for a generally higher density of PSs per tile. Specifically, nearly 70% of the tiles contain at least 4 MPs. However, approximately 20% of the tiles still exhibit a number of MPs equal to 0 or 1, which is insufficient for computing the standard deviation. On the other hand,