PSI - Issue 44

Marialuigia Sangirardi et al. / Procedia Structural Integrity 44 (2023) 1602–1607 M. Sangirardi et al./ Structural Integrity Procedia 00 (2022) 000 – 000

1606

5

and a 4th order Butterworth high-pass filter at 0.1 Hz cut-off frequency, and processed to calculate the Normalized Power Spectral Density. Figs. 4a-b report, in the top right corner, the orientation of the instrument. Observing the recording in the x direction (Fig. 4a) three frequencies clearly emerged, which are equal to 4.8 Hz, 5.9 Hz, and 15 Hz, which were consistent with the expected natural frequency of the tank for a translational mode. The processing of the signal in y direction, instead, only showed one frequency equal to 13.4 Hz. Due to the lack of symmetry in these outcomes, in spite of the geometrical symmetry of the structure, these results were considered insufficient, by themselves, to fully characterize the dynamic behaviour of the elevated water tank. The identification of the dynamic properties of the tank from accelerometers resulted incomplete. The application of the computer-vision based methodology described above complemented the previously mentioned results. Videos were recorded at 30 fps with a commercial camera having 24.3 MPx resolution from the ground level. Two different fields of view were considered, with their axis being parallel to y and x, so to have results that could be interpreted in the x and y direction, respectively.

Fig. 4. Frequencies detected from accelerometric recordings in x (a) and y (b) directions. Results obtained through the computer-vision-based method with field of view oriented parallel to y (c) and x (d). The results obtained from the computer-vision-based method are reported in Figs. 4c-d, showing good agreement with those obtained from the accelerometric signals, proving that the proposed contact-less technique is able to identify with fairly good precision the same natural frequencies provided by the routine post-processing of accelerations. The percentage error resulted 2÷5% for the first two frequencies and might be related to environmental conditions such as light and movement which, in field applications, are expected to affect the results much more than in the laboratory were a much higher control can be ensured. More interestingly, these results also point at some features of the structural behaviour of the tank that were not evident in the accelerometric signals, possibly because of the resolution of the instruments. Specifically, both points of view (that is different recordings) allowed to identify two frequencies, the first being between 4.5 Hz and 4.9 Hz, and the second between 6.0 Hz and 6.2 Hz. Pairing this observation with the outcomes of the numerical simulations allowed to clarify that the first two frequencies detected also by the accelerometer are representative of a torsional and two uncoupled translational modes of vibrations, which are in line with the symmetry of the structure. This interpretation is also compatible with the position and the orientation of the accelerometer. 5. Conclusions A computer-vision-based method was used to detect the natural frequencies of a reinforced concrete elevated water tank situated in Rome. Videos were recorded by a commercial digital camera and processed by analysing the variation of grey intensity and the motion of a set pixels. The two methods provided consistent results, being the tracking of intensity slightly more accurate in providing the actual values. In both cases, motion magnification was necessary for improving accuracy, given the small amplitude of the oscillations caused by environmental noise.