Covariance driven SSI Data driven SSI (UPC) mode f (inHz) ˜σf d (in%) ˜σd ˜σϕmax f (inHz) ˜σf d (in%) ˜σd ˜σϕmax 1 4.039 0.34 1.1 23 13 4.031 0.21 1.3 22 22 2 6.292 0.19 0.6 52 27 6.282 0.11 0.7 19 22 3 9.730 3.42 3.0 92 47 9.872 1.17 2.1 32 46 4 13.19 0.67 1.3 73 31 13.31 0.38 1.4 22 48 5 15.72 0.70 1.8 20 32 15.73 0.39 1.8 19 39 Table 2 Overview of the estimated first 5 modes of S101 Bridge with natural frequencies f , their relative confidence bounds ˜σf =σf /f ·100, the damping ratios d, their relative confidence bounds ˜σd =σd/d·100 and the relative confidence bounds of the mode shape element of maximal amplitude ˜σϕmax =σϕmax/ϕmax · 100. To summarize the system identification results from covariance and data driven SSI, it can be said that both approaches give practically identical estimates in this test case when taking the obtained confidence bounds into account. However, the data driven approach seems to yield frequency and damping estimates that have lower confidence bounds than the covariance driven approach. Confidence bounds for the mode shape estimates are comparable for both approaches. Confidence bounds are low on modes that seem to have stabilized in the stabilization diagram (e.g. on modes 1 and 2 at model order 40), while they are high on modes that have not stabilized yet (e.g. mode 3 at model order 40). Confidence bounds on frequency estimates are very low (lower than 1% on stabilized modes), while they are much higher on damping and mode shape estimates. 4.5 Monitoring during Progressive Damage Test During the progressive damage test of the S101 Bridge, more than 700 datasets were available. Some of them contained erroneous data due to destruction work on the bridge or other influences that were deleted. On the left 680 datasets, an automated monitoring procedure was applied, that did the system identification and confidence interval computation automatically for each dataset. This means, that for each dataset a stabilization diagram was built with the SSI algorithms, containing model orders from 10 to 70. Then, the modes were chosen automatically using stabilization criteria such as thresholds for the damping estimates and confidence interval bounds, small frequency deviation between successive model orders, a minimum number of appearances of a frequency in the diagram and the MAC value between successive model orders. The results of the frequency monitoring of all datasets are displayed in Figure 6 and the respective damage scenarios are explained in Table 1. Especially the frequency drop can be clearly seen when one column of the bridge was lowered before it was lifted up again (between A and G). This affected mainly the second, third and fourth mode, while the frequency changes in the first mode were less important. EsConfidence Intervals of Modal Parameters during Progressive Damage Test 247
RkJQdWJsaXNoZXIy MTMzNzEzMQ==