p+1=q=11 at system orders n=1,...,70. Confidence bounds were obtained by cutting the data into 100 blocks. In Figure 4, the stabilization diagrams of the natural frequencies from both SSI methods are presented, where the confidence interval of each frequency is plotted as a horizontal bar. The obtained confidence bounds on the frequencies were used to clean the diagrams: Modes with frequencies having big confidence bounds are likely to be spurious and are erased. In this case, all modes with confidence bounds bigger than 2% of the frequency value were deleted. Fig. 4 Stabilization diagrams with covariance driven SSI (left) and data driven SSI (right) containing confidence intervals on the frequencies (top: full diagrams, bottom: zoom on first mode). From the stabilization diagrams, the modes of the system are chosen. In Table 2, an overview of the obtained modal parameters and their confidence bounds at model order 40 is given. Note that all confidence bounds are relative values in percent, i.e. the standard deviation of a value divided by the value and multiplied by 100. For the mode shapes, only the relative confidence bound is displayed for the mode shape element of maximal amplitude. Finally, the obtained mode shapes at the 14 sensors of one side of the bridge deck are displayed with their confidence bounds in Figure 5. From the 15th sensor on the other side of the bridge deck (see also Figure 2), information about the kind of the mode is obtained. So, modes 1, 3 and 5 are vertical bending modes and modes 2 and 4 are torsional modes. Confidence Intervals of Modal Parameters during Progressive Damage Test 245
RkJQdWJsaXNoZXIy MTMzNzEzMQ==