PSI - Issue 44

Gianluca Standoli et al. / Procedia Structural Integrity 44 (2023) 2066–2073 G. Standoli et al. / Structural Integrity Procedia 00 (2022) 000 – 000

2069

4

connected in chain, assuring the synchronization of the measures. Data are acquired for 20 minutes every hour, with a sampling frequency of 200 Hz. When the continuous acquisition process stops, acquisition on trigger activates and, in case of events, registrations start with sampling frequency of 1000 Hz and duration of 90 seconds of pre- and post-triggering. Environmental data, concerning external temperature, external humidity, and wind speed, are collected through a weather station positioned in the proximity of the structure, available for consultation on the site http://app.protezionecivile.marche.it/sol/info.sol?lang=it.

Fig. 3. Sensors layout: respectively the continuous monitoring system (in green) and the short-time monitoring system (in blue).

To assess the initial conditions of the towers at the start of the long-time monitoring process, a short-time acquisition has been made, using 18 monoaxial piezoelectric accelerometers (Fig. 3), with a sensitivity of 10 V/g, fixed in groups of three sensors in six corners of the bell cell (in proximity of the MEMS sensors). This acquisition lasted 40 minutes with a sampling frequency of 1000 Hz. 3.2. Modal identification Data are elaborated automatically through a self-made script developed in Matlab© environment. The towers data are analysed separately, to highlight the difference in dynamic behaviour including the influence of temporary (safety provisions) reinforcement interventions. The script firstly pre-processes the short-time acquisition, applying de-trending, low-pass filtering and then decimating data in the range of 0-12.5 Hz, which is the one of interest for this type of structure. Then modal parameters identification is operated through automatic Enhanced Frequency Domain Decomposition (EFDD) (Jacobsen et al. (2006)) and Stochastic Subspace Identification - Covariance Based (SSI-Cov) methodology (Peeters and De Roeck (1999)) for both towers, and the resulting modal frequencies and mode shapes (Table 1) become the targets for modal tracking over the continuous monitoring data.

Table 1. Comparison between target and median automatically identified parameters.

̅̅̅

̅̅̅ 2021 [Hz] 1.719

Mode

Left Tower

Right Tower

Comparison

2021 [Hz] 1.805

Mode Shape

Mode shape

Δ f [%]

1 2 3 4

Translational Y Translational X Bending X + Tors

Translational XY

5.00 4.51 0.71 0.99

2.318 4.228 5.906

2.218 4.198 5.848

Translational X + Torsional

Bending X + Torsional Bending Y + Torsional

Bending XY + Torsional

From the identification operated over the first dataset, it is observable that the two towers, even though they may appear similar in terms of modal frequencies values, they exhibit very different dynamic behaviours, with mode