Safety Detection of Bridge Structure Using Refined HHT and Fuzzy Regression Approach

• Main Authors: Chih-Wei Huang, 黃志偉
• Other Authors: Jeng-Wen Lin
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/93290878500566427165

博士 === 逢甲大學 === 土木暨水利工程博士學位學程 === 98 === This thesis presents a refined Hilbert-Huang transform (HHT) and the fuzzy regression approach for safety detection of bridge structure, mainly with three contributions: (1) has developed a refined HHT and tested a simulated signal to prove its efficiency; (2) has applied the refined HHT to a real bridge structure to estimate the vibration frequency in axial and cross directions of bridge; (3) has used maximum displacement data of the bridge under various earthquakes and adopted a fuzzy regression approach so as to establish the bridge “alert-action level” evaluation principle. First, a simulated signal was tested, where each mode of the signal would be separated through a designed stopband bandpass filter and modal analysis would be carried on each mode independently. The modal analysis procedure is as follows: the separated modal signal will obtain one of several intrinsic mode functions (IMFs) via empirical mode decomposition (EMD). Through selecting the largest one of the correlation coefficient between each modal signal and its IMFs as the modal signal, the remaining IMFs would be regard as noise and deleted, then the marginal spectrum would be used to estimate each modal frequency. Using such a selection process, each modal frequency of the signal can be evaluated accurately. Second, the refined HHT approach was applied to the real bridge structure to estimate the vibration frequency of the bridge. This study takes Mao-Luo-Hsi Bridge in Nantou County as an example, carries on ambient vibration test in sections E and F of the bridge separately, and measures velocity responses in axial and cross directions of the bridge. Finally, the bridge SAP2000 model was established according to the evaluated vibration frequency in order to conduct the modal signal analysis of the considered sections of the bridge, in which the maximum displacement data were generated subjected to earthquakes. Subsequently, one to seven magnitudes of earthquake intensities would be tested to obtain maximum displacement data of the bridge surface. Due to the uncertainty of measurement, this study utilized the fuzzy regression approach to establish the bridge “alert-action level” evaluation principle using the maximum displacement data of E and F bridge surfaces. The action level was determined by the maximum earthquake intensity corresponding to the slope of the fuzzy regression line being relatively large (e.g. larger than 0.5), while a lower level of the earthquake intensity was determined as the alert level. In comparison with the theoretical value, the proposed refined HHT has provided more stable and accurate results than the traditional FFT. Hence, it has been applied to a real bridge structure to establish the bridge “alert-action level” evaluation principle. Such “alert-action level” evaluation principle helps preventing disasters caused by earthquakes and protecting lives and properties.