| 摘要: | 在結構健康監測時，常使用模態參數來辨別訊號的不同，在過去分解訊號時大多是使用經過傅立葉轉換(Fourier Transform)後的頻譜進行分析，此方法只適用於穩態的分析，即使後來的學者提出短時距傅立葉轉換(Short-Time Fourier Transform, STFT)，此法利用窗函數來獲得該時刻的頻譜分析，但仍然無法得到訊號的瞬時頻率，而且會因為窗函數(window function)選擇的不同造成分析結果的差異。
為了處理地震力所造成的非穩態訊號所以使用黃鍔博士提出的希爾伯特-黃轉換(Hilbert-Huang Transform, HHT)的方式處理非穩態和非線性的訊號，解決傳統傅立葉轉換需要預設基底函數和無法得到瞬時的頻率的缺陷。透過希爾伯特-黃轉換將訊號展開至時頻域並且觀察訊號瞬時頻率的變化，量化出兩個模態參數II和NI來進行結構健康監測的探討。
本研究使用上述的訊號分解方法搭配有限元素軟體ABAQUS建立局部斷面折減的模型所輸出的加速度歷時資料，繪製出訊號隨時間變化的時頻譜，在時頻譜上擷取資料計算出模態參數，根據模態參數的趨勢變化來顯示出訊號輸入與輸出時的不同，藉由這個方式來比較當結構發生局部斷面折減時，模態參數可能的變化情況，從研究結果可以發現當局部斷面折減程度越大時II會下降，NI值會上升，並且透過頻率折減率和DN值判斷局部斷面折減可能發生位置，並且找出其不足之處來做為未來使用此方法進行結構健康監測時的一種參考。
;When it comes to Structure Health Monitoring, we always use the modal parameters to recognize the difference of the output signal. In the past, we used Fourier Transform to decompose the signal, and analyze the frequency spectrum. However, this method may only be suitable to stationary and nonlinear signal. Even though some researcher presented an alternative method called short-time Fourier Transform (STFT) to solve the disadvantage of Fourier Transform, we are not able to know the instantaneous frequency in the signal.
In order to solve the signal involved nonlinear and nonstationary property (earthquake), we used Hilbert-Huang Transform method to deal with it. This method was developed by Dr N, E. Huang in 1998. There are two advantages in this method. First, we don’t need to set the basic function. Second, we can identify the change of the instantaneous frequency. Through the Hilbert-Huang Transform, the signal can be expanded to the distribution of energy in time-frequency domain. We constructed two modal parameters (II,NI)in the time-frequency spectrum to quantify the physical behavior of the structure and use the parameters to monitor the health of the structure.
In this study, we used a finite element software; ABAQUS, to construct a five-floor structure model which has a local damage. Then, used the acceleration from the top of each floors as the original signal to draw the time spectrum. We selected the data from the spectrum and converted it into two modal parameters (II,NI) and a new parameter (DN) to compare the differences of each model. Moreover, to find the relation between the different models and apply this method to the real case of structure health monitoring. |
