PSI - Issue 84

Antonio S. López-Cuervo et al. / Procedia Structural Integrity 84 (2026) 223–230

228

ambient temperature at the time of each test, inferred from the daily minimum and maximum temperatures and their corresponding times of occurrence, as provided by the Cartuja (5515X) meteorological station of the Spanish Agency of Meteorology (AEMET, 2025). As shown, AVT-SG-1 was conducted at the lowest estimated temperature (6.7 °C), whereas AVT-SG-2 took place at the highest one (27.8 °C). Consistently, these tests provide, respectively, the highest and lowest frequency estimates for the identified modes. Figure 6 plots the identified frequencies for each mode across AVTs (adimensionalized by their mean value) against the temperature estimates, showing the negative correlation between frequency and temperature. This behavior is consistent with previous observations reported for bridges (Xia et al., 2012).

Table 2. Maximum and minimum temperatures data from the Cartuja (5515X) meteorological station, together with values of temperature estimated at the hour of the test. An asterisk (*) denotes that the hour value was inferred from the previous day because the corresponding data were not available. AVT-Acc-1 AVT-Acc-2 AVT-SG-1 AVT-SG-2 T min [°C] 8.6 -0.7 6.5 14.5 Hour T min 6:58 6:33 5:54* 5:26 T max [°C] 21.2 12.8 21.8 30.5 Hour T max 13:28 14:50 14:55 13:58 Hour test 13:17 14:28 10:55 12:31 T estimated [°C] 20.8 12.2 6.7 27.8

Finally, a good agreement is observed between the identified frequencies and the FEM predictions for Modes 1 and 3, with deviations below 5% with respect to the experimentally identified values. In contrast, Mode 2 exhibits a substantially larger discrepancy, with the FEM predicting 5.458 Hz compared to the mean identified frequency of 4.100 Hz.

Fig. 6. Influence of temperature on identified frequencies across different AVTs. Frequencies are normalized mode-wise by the mean frequency over all AVTs

3.2. Linear independence of identified modes

The linear independence of the identified modes is discussed next to assess the performance of the EfI algorithm. For brevity, the analysis is limited to AVT-Acc-1 and AVT-SG-2. Figure 7(a) shows the AutoMAC matrix of the modes identified in AVT-Acc-1. Modes 1 (first bending-X) and 3 (first bending-Y) are nearly orthogonal (MAC close to zero), whereas Modes 1 and 2 (second bending-X) exhibit a