PSI - Issue 62

Fabio Micozzi et al. / Procedia Structural Integrity 62 (2024) 848–855 Author name / Structural Integrity Procedia 00 (2024) 000 – 000

854

7

To analyse the frequency contents of the readings of the displacement transducer (DT), video camera (FVB), accelerometer at midspan (AC), accelerometers on the video camera tripod (ACTV and ACTH in the vertical and horizontal directions respectively), their Power Spectral Density (PSD) is shown in Figure 6. It is observed that only AC and DT permit to recognize the subsequent two modal frequencies. It is also noted that FVB is affected by pronounced noise between 12 and 32 Hz; following the analysis of the frequency contents of ACTV and ACTH, it can be deduced that such noise originates from the ground vibrations induced by the vehicular traffic, filtered by the tripod, and transmitted to the video camera. This noise has a frequency content outside the range of the first vibration mode of the bridge being monitored. Nevertheless, it has a negative influence of the accuracy of the displacement time histories because of the high-frequency spikes, as observed in Figures 3, 4, and 5.

Fig. 6. Power Spectral Density from the displacement transducer (DT), the video camera (FVB), the accelerometer at midspan (AC), and the accelerometers on the tripod of the video camera (vertical ACTV and horizontal ACTH directions). 4. Conclusions This study illustrated the experimental results obtained using a cost-effective vision-based structural monitoring system to an existing post-tensioned concrete bridge under normal service conditions. The interest for such bridge typology is given by its large diffusion in the road network in Italy and many other countries in conjunction with the fact that previous applications of vision-based structural monitoring documented in the technical literature mostly involved more deformable steel bridges and footbridges. The considered bridge was selected because of the limited distance from the ground of the deck at midspan, allowing the installation of a displacement transducer as reference measure. The video camera was placed under the bridge deck, close to the abutment, pointing at a target fastened at midspan. In this way it is possible to measure the vertical and transverse movement of the bridge deck in a position protected from rain and direct sunlight. A simple computer vision processing algorithm was adopted, allowing subpixel accuracy in the identification of the displacement of the target within a selected portion of the video. The video processing was made in real-time, thus, it was possible to visualize and store the extracted displacement as they were measured, without the need to save large memory-consuming videos for subsequent processing.