**4.2. System identification of the bridge**

Before the identification, data processing on the acceleration data is implemented. Firstly, base-line correction is performed to get rid of offset value. The next step is to remove linear trends from the data using de-trend technique. Besides, the mean value (DC) of the data

**Figure 7.** Identification of strong wind: meteorology data and SHM data.

**Figure 8.** Identification of wind direction from the deck mid-span SHM data.

is subtracted. In order to remove the unwanted noise component, a standard fourth order Butterworth band-pass filter with the first corner frequency of 0.05 Hz and the second corner frequency of 5.0 Hz is performed [21]. For all these works, MATLAB computing program developed by the Math Works [22] is utilized. For more refined results from FFT (Fast Fourier Transform), data averaging technique including windowing and overlapping is also implemented. Window length is determined considering the minimum frequency range of the modal values obtained from the previous experimental study for the bridge in literature.

obtained for "Before" range since the data is relatively distorted with traffic noise. Therefore, "After" range is considered for the comparison with "During" range. The obtained SHM data from the accelerometers mounted at the specific points of the bridge are utilized for structural identification of the bridge. The locations of the accelerometers are indicated with the considered directions in **Figure 10**. In this figure, the FFT analyses of the all accelerometers are also given for "During" range to determine modal characteristics of the bridge. This effort is repeated again for "After" range and the outcomes from frequency-domain analysis are summarized in **Table 2**. As mentioned previously, the results from "After" range can be considered as natural vibration characteristics of the bridge. In addition to the other studies in literature, "After" range provides an opportunity to make calibration of the developed FE model. Therefore, the comparison of the outcomes from "After" with those from the developed FE model is made as

Structural Identification (St-Id) Concept for Performance Prediction of Long-Span Bridges

http://dx.doi.org/10.5772/intechopen.71558

53

**Figure 9.** Three ranges on SHM data: "Before", "During" and "After".

According to % change in **Table 3**, the maximum error % in terms of frequency is obtained for Mode-2 with approximately 12%. For the other modes, the error is in the range of 0–3.0%.

given in **Table 3**.

In order to determine the effects of the strong wind on dynamic characteristics and operational performance of the bridge, all SHM data are divided into three ranges as shown in **Figure 9**: "Before", "During" and "After". The data analysis shows that meaningful results are not Structural Identification (St-Id) Concept for Performance Prediction of Long-Span Bridges http://dx.doi.org/10.5772/intechopen.71558 53

**Figure 9.** Three ranges on SHM data: "Before", "During" and "After".

is subtracted. In order to remove the unwanted noise component, a standard fourth order Butterworth band-pass filter with the first corner frequency of 0.05 Hz and the second corner frequency of 5.0 Hz is performed [21]. For all these works, MATLAB computing program developed by the Math Works [22] is utilized. For more refined results from FFT (Fast Fourier Transform), data averaging technique including windowing and overlapping is also implemented. Window length is determined considering the minimum frequency range of the modal values obtained from the previous experimental study for the bridge in

**Figure 7.** Identification of strong wind: meteorology data and SHM data.

**Figure 8.** Identification of wind direction from the deck mid-span SHM data.

In order to determine the effects of the strong wind on dynamic characteristics and operational performance of the bridge, all SHM data are divided into three ranges as shown in **Figure 9**: "Before", "During" and "After". The data analysis shows that meaningful results are not

literature.

52 Bridge Engineering

obtained for "Before" range since the data is relatively distorted with traffic noise. Therefore, "After" range is considered for the comparison with "During" range. The obtained SHM data from the accelerometers mounted at the specific points of the bridge are utilized for structural identification of the bridge. The locations of the accelerometers are indicated with the considered directions in **Figure 10**. In this figure, the FFT analyses of the all accelerometers are also given for "During" range to determine modal characteristics of the bridge. This effort is repeated again for "After" range and the outcomes from frequency-domain analysis are summarized in **Table 2**.

As mentioned previously, the results from "After" range can be considered as natural vibration characteristics of the bridge. In addition to the other studies in literature, "After" range provides an opportunity to make calibration of the developed FE model. Therefore, the comparison of the outcomes from "After" with those from the developed FE model is made as given in **Table 3**.

According to % change in **Table 3**, the maximum error % in terms of frequency is obtained for Mode-2 with approximately 12%. For the other modes, the error is in the range of 0–3.0%.

eling of the Bosphorus Bridge are estimated well and are properly implemented. Accordingly,

**FEM Experimental Change (%)**

**Period [s] Freq. [Hz] Period [s] Freq. [Hz] Period [s] Freq. [Hz]**

Structural Identification (St-Id) Concept for Performance Prediction of Long-Span Bridges

http://dx.doi.org/10.5772/intechopen.71558

55

After the destructive earthquakes in last 2 decades in Turkey, Izmit (1999) and Duzce (1999) earthquakes, the public awareness of structural earthquake safety and performance of the existing structures in Turkey has increased progressively. General Directorate of Turkish State Highways (KGM) conducted a number of rehabilitation projects (JBSI) [23] for the most critical long-span bridges in Turkey, such as the Bosphorus Bridge. The related studies for the Bosphorus in literature were basically focused on the uniform support earthquake analysis (U-sup) of the bridge. Therefore, the multi-point earthquake analysis (Mp-sup) is required to better understand the seismic behavior of the Bosphorus Bridge. Considering these recommendations, this chapter aims at determining the effects of spatially varying earthquake motion on

the developed FE model can be reliably utilized for advanced analysis of the bridge.

the Bosphorus Bridge using the calibrated FEM of the bridge in the previous section.

In order to simulate site-specific ground motions, the geographic coordinates of the bridge's support points have firstly to be determined. As indicated in **Figure 11**, the support coordinates of the bridge are obtained depending on the general coordinates of the bridge. **Figure 11** also presents the general considerations of the multi-point earthquake analysis of the bridge. Taking the scenario earthquake of Mw = 7.4 predicted to occur with the probability of 70% in next 30 years in Istanbul and these coordinates of the bridge into consideration, the stochastic modeling technique proposed in [24] is used to generate spatially varying site-specific earthquake ground motions.

The simulation process is performed and the acceleration ground motion time-histories (ATH) are generated for the Bosphorus Bridge. Although the process yields to the ATHs, the displacement ground motion time histories (DTH) need to be obtained for the multi-point earthquake analysis. Therefore, the DTHs are presented in **Figure 12** instead of the ATHs. As shown in **Figure 12**, the triple-direction (two horizontals and one vertical) ground motions are generated for the each considered multi-point, A, B, C and D. Total number of twelve ground

motions are defined for the analysis.

**Mode number Mode shape**

**5. Utilization, decision-making and performance prediction**

**Table 3.** Comparison of FEM with those from experimental results.

**FE model and experimental**

Mode-1 1st Lsym 12.984 0.077 12.766 0.078 1.707 −1.679 Mode-2 1st Vasym 7.217 0.139 8.065 0.124 −10.505 11.738 Mode-3 1st Vsym 6.453 0.155 6.250 0.160 3.251 −3.149 Mode-4 1st Lasym 4.926 0.203 4.967 0.201 −0.819 0.825 Mode-5 2nd Vsym 4.561 0.219 4.525 0.221 0.809 −0.803 Lsym: Lateral symmetric; Lasym: Lateral asymmetric; Vsym: Vertical symmetric; Vasym: Vertical asymmetric; T: Torsional.

**Figure 10.** The FFT analysis of the extreme wind data for "During" range.

Mode-2 is identified only from "During" range instead of "After" range considered for ambient vibration. Therefore, such high error in Mode-2 is dependent on that the mode is not be identified from "After" range. Generally, the error range for the first five mode shapes of the bridge is allowable level. These conclusions reveal that the considerations utilized in FE mod-


**Table 2.** FFT analysis results for "During" and "After" ranges.


**Table 3.** Comparison of FEM with those from experimental results.

Mode-2 is identified only from "During" range instead of "After" range considered for ambient vibration. Therefore, such high error in Mode-2 is dependent on that the mode is not be identified from "After" range. Generally, the error range for the first five mode shapes of the bridge is allowable level. These conclusions reveal that the considerations utilized in FE mod-

**During After Change (%)**

**Period [s] Freq. [Hz] Period [s] Freq. [Hz] Period [s] Freq. [Hz]**

**Frequency/period [Hz]/[s]**

Mode-1 1st Lsym 14.706 0.068 12.766 0.078 15.196 −13.191 Mode-2 1st Vasym 8.065 0.124 7.853 0.127 2.698 −2.362 Mode-3 1st Vsym 6.096 0.164 6.250 0.160 −2.469 2.531 Mode-4 1st Lasym 4.854 0.206 4.967 0.201 −2.265 2.318 Mode-5 2nd Vsym 4.561 0.219 4.525 0.221 0.798 −0.792

Lsym: Lateral symmetric; Lasym: Lateral asymmetric; Vsym: Vertical symmetric; Vasym: Vertical asymmetric.

**Figure 10.** The FFT analysis of the extreme wind data for "During" range.

**Table 2.** FFT analysis results for "During" and "After" ranges.

**Mode number Mode shape**

54 Bridge Engineering

eling of the Bosphorus Bridge are estimated well and are properly implemented. Accordingly, the developed FE model can be reliably utilized for advanced analysis of the bridge.
