依據ATC-40及FEMA-273以層間變位角作為結構物之損傷指標。本文採用國家地震中心所發表之兩組震動台試驗，其震動台分別輸入Kobe及El Centro地震力，並將數據藉由加速度二次積分及測得的真實位移計算層間變位角，得知層間變位角無法看出結構物降伏損害位置。 因此，本文另外採用HHT方法探討結構物損傷偵測之可能性，並定義頻寬比(RB)作為損傷指標。將震動台數據進行HHT及FFT分析並計算頻寬比，藉由頻寬比進一步了解結構物損傷情形。 研究成果顯示： (1)當結構物在線彈性反應時，頻寬比在HHT邊際譜與FFT頻譜中僅有微小改變。 (2)當結構物進入非線性反應後，觀察HHT邊際譜發現頻寬比有明顯增加之情形，藉此可獲得結構物的降伏發生。相反的，FFT頻譜無法看出此現象。 (3)頂樓之頻寬比變化相較於其它樓層明顯。 (4)分析頂樓之加速度反應可偵測結構物是否發生損害。 因此，將HHT邊際譜進行平滑處理所得之頻寬比運用於結構物損害中，是項成功的損傷指標。 Referring to the ATC-40 and FEMA-273, the interstory drift is used as a damage detection index. This research uses two benchmark models which built by the National Center for Research on Earthquake Engineering (NCREE). Shaking table test data from benchmark models subjected to adjusted Kobe and El Centro earthquakes are analyzed to evaluate interstory drift using the acceleration integration method and measured displacement (LVDT). The results of the yielding point in interstory drift curve are difficult to find out when member damage occurs. Therefore, this study investigates the feasibility of detecting structural damage using the HHT method and the Ratio of Bandwidth (RB) is proposed as the damage detection index. Shaking table test data are analyzed to evaluate the RB using the Hilbert-Huang Transform (HHT) and the Fast Fourier Transform (FFT) methods, respectively. The result of this study shows that: (1)When the response of the structure is in the elastic region, there is very small change in the RB value from the HHT spectra and the FFT spectra. (2)The incremental change in RB estimated from the HHT spectra versus the PGA value can be seen when the structure response in nonlinear i.e., member yielding occurred, but not in the RB from the FFT spectra. (3)The RB of the top floor reveals the highest change value than other floors. (4)Structural damage can be detected using only the acceleration response data from the top floor. Therefore, the ratio of bandwidth (RB) which estimated from the smoothed HHT spectra is an effective and sensitive damage index for the detection of structural damage.
