PSI - Issue 78

Andrea Pozè Falet et al. / Procedia Structural Integrity 78 (2026) 325–332

328

3. The 2024 Experimental Campaign The 2024 campaign of field tests lasted for three whole days on February 7 th , 8 th , and 12 th . All simply supported spans and all piers were monitored; here, only data from the deck are reported. These were investigated on the third day, using 39 channels, all oriented vertically and upward, as depicted in Figure 4 (the number was limited by the DAQ system's maximum capability). These were obtained using 14 393B31 uniaxial piezoelectric accelerometers (sensitivity: 10 V/g), 20 PCB 3701G3FA3G uniaxial capacitive accelerometers (sensitivity: 1 V/g), and five PCB356B18 triaxial piezoelectric accelerometers (sensitivity: 1 V/g). Thanks to the relatively high flexibility of the deck, the output amplitudes were always large enough to be easily detectable, even with lower-sensitivity sensors. Three acquisitions, or ‘runs’, were performed: Run 0 (lasting 26’47”), Run 1 (10’31”), and Run 2 (26’10”). Run 1 was discarded due to its limited duration; among the remaining two, Run 2 was then chosen for the following analyses reported hereinafter.

Fig. 4. Sensor layouts of the 2024 campaign (each span was investigated separately), seen looking in the stream direction (light blue arrow).

3.1. Results of 2024 Ambient Vibration Testing Here, the results of the most recent experimental campaign are briefly reported. For brevity, neither the specific System Identification algorithm (SSI-DATA) nor the used DSFs will be detailed any further. For SSI-DATA, one can refer to the seminal work of (Van Overschee & De Moor, 1996). Figure 5 shows an example of acquired signals, considering all output channels, in the time domain. In this specific case, the acquisition lasted for circa 45 minutes, such that the assumptions of White Gaussian Noise input and ergodic output signals were respected. The other acquisitions lasted for similar durations as well. In all cases, a sampling rate of = 256 Hz was used. Thus, due to the Nyquist limit, the observable range is included from 0 to 128 Hz, which was proved to be amply sufficient for the main modes of the deck.

Fig. 5. Example of acquired signals from Span C3.

These identifications led to the results showcased in Figure 6, using one instance of the channels on the third span (C3) as an example. In fact, it should be stressed that all spans are nominally identical and separated, as simply supported over the piers. Furthermore, the AVTs did not exhibit any hint of joint locking; thus, even if acquired and inspected simultaneously, they can be considered as five independent structures, allowing for direct comparison among them.