PSI - Issue 64

F.-B. Cartiaux et al. / Procedia Structural Integrity 64 (2024) 285–292 Cartiaux / Structural Integrity Procedia 00 (2019) 000 – 000

289

5

4.2. Massive spectral decomposition and cycle range analysis The first way to process the large volume of monitoring data is to extract features at a regular period like one day or one week (or more), which will represent the characteristics of the response of the bridge decks to the traffic loads by gathering results from several events on the same period. Two of these features are analyzed here: • The spectral decomposition of the signal obtained by a simple Fast Fourier Transform (FFT) on a fixed 10 seconds rolling window on all events of each day, • The number of measured cycles classified by their range by applying a Rainflow counting method on all events of each week. The last is usually used for fatigue assessment, but in this case, we consider it as a synthetic weekly feature, without fatigue consideration, as it still describes a meaningful property of the response of the bridge (strain or acceleration range) that can be used to check ageing effects and stability over the long term. These features are illustrated on the example of vertical acceleration (for the spectral decomposition) and longitudinal strain (for the cycle range), displaying the whole results over the two years of monitoring. 4.2.1. Spectral decomposition As an example, the heat map of the spectral decomposition of the vertical acceleration on the first (end) span of the upstream deck at one third of its length from the first pier is given here-below (Fig. 5).

Fig. 5. Spectral decomposition of the vertical acceleration of the upstream deck, first span, at one third of the length from the first pier.

Three very clear peaks with frequency values stable over the two years are identified at 6.1 Hz, 12.6 Hz and 18.5 Hz. In addition, the domain comprised between 15.2 Hz and 15.8 Hz concentrates the most energy. On the location at mid-span (not displayed on Fig. 5), another domain shows this concentration at lower frequencies between 4.5 Hz and 6 Hz. Also, on the second span (not displayed on Fig. 5), the three clear peaks are the same, but the two smoother energy domains are at lower frequency (as the span is longer) around [3.9 Hz, 5.1 Hz] at mid-span and [12.1 Hz, 13.3 Hz] at the third of the span length. We can expect that these peaks and domains are related to modal properties of the bridge deck, including bending and torsion modes but also local modes coupled with the vibrations of the external tendons inside the box girders. Without going further into this explanation at this stage of the study (this would require a model of the bridge, which is foreseen), this analysis already allows to set a first criterion for the health assessment of the deck structure, by checking the stability of the frequency of the peaks and domains with the most energy over the whole duration of the continuous monitoring, with one result every day.