利用小波轉換技術於結構振動訊號之解析

Application of Wavelet Transforms to System Identification & Damage Detection in Frame Structure

辜琪媜(Chi-Jen Ku)

指導教授 ： 羅俊雄

國立臺灣大學/工學院/土木工程學研究所/碩士(2005年)

https://doi.org/10.6342/NTU.2005.00027

摘要

在監測動態結構訊號時，可以利用FT(Fourier-based analysis)來分析訊號在頻率域情況下的資料，但是，所分析的訊號要以為線性及穩定的前提下才能進行分析，這對於訊號若有發生異常情況下的觀測極為不利，因此，可以利用WPT(Wavelet Packet Transform)在時間域及頻率域兩者擇一的情況下進行分析。在此篇論文中運用的WPT取得了最佳的訊號後，且進一步的作了以下的分析：結構物破壞情形的識別以及系統特性的識別。
進行結構物破壞情形的識別時，所要分析的訊號是經過WPT處理以及能量特性的篩選後所得到的，而且所要分析的訊號一定有包含了訊號微弱異常的歷時部份，在將其作進一步的分析；首先，可以求得CWCS(Cross-Wavelet Coefficient Spectrum)以及相角的變化後，進一步可以藉由觀察CWCS的最大值的振幅以及相角的延遲情況的也可得到破壞所會產生得特性，或者是利用Holder Exponents來得到訊號產生不連續的破壞情形。而在進行系統特性的識別時，所要分析的訊號也是要先經過WPT的篩選後，並且要滿足特定的能量值以及Entropy Index後，才能加以分析；接著，則可以運用HT(Hilbert Transform)的特性可以得到系統的瞬間頻率以及阻尼的變化情形來觀察系統的特性。
將門形構架放於振動台上後，並且引入地震歷時進行試驗，所收集到的訊號則可以藉由以上所提過的方法進行分析。而所收集到的訊號，則可以藉由WPT的處理後，高頻的部分則可以觀測出發生異常的情形(所指的是門形構架發生了勁度損失或是有降服的情況發生)；低頻的部分則可以進行門形構架的系統識別。

關鍵字

小波轉換之應用分析 ； 結構物破壞情形的識別 ； 系統特性的識別

並列摘要

Condition monitoring of dynamic system based on vibration signatures has relied on Fourier-based analysis as a mean of translating vibration signals in the time domain into the frequency domain. However Fourier analysis provided a representation of linear and stationary signal only. It is difficult to detect the signal pattern with abnormal and diluted information. The wavelet packet transform (WPT) is used as an alternative means of extracting time-frequency information from vibration signature. In this paper two research directions by using WPT to extract features from vibration signal are developed：one is the damage detection and the other is the modal parameters identification.
For damage detection the WPT feature extraction techniques are introduces to detect the occurrence of damage during the structural response time history. In the beginning based on feature vector of node energy of WP analysis, significant decomposed signals are developed. The amplitude of cross-wavelet coefficient power spectrum and the instantaneous phase features are generated. Damage detection can be identified from the maximum value of cross-wavelet coefficients and the phase lag. The holder exponents are also incorporated to detect the discontinuity of the signals. These methods can provide features of abnormal signal. To identify the modal parameters (natural frequencies and damping ratios) the wavelet packet sifting process is used to extract the dominant vibration signal of the system response. The sifting process can sift the decomposed signals under the conditions of a specified component energy and the entropy index. By converting the analytical signal using the Hilbert transform and its instantaneous frequency, the frequency-time domain of modal frequency and damping can be identified.
Application of the above-mentioned methods to data collected from a collapse test of a portal frame subject to seismic excitation on the shaking table is conducted. Through WPT on high frequency signal the occurrence of time where abnormal situation along the time response (such as time of negative stiffness in the restoring force, time of maximum yield force occurred during the excitation) can be detected. Dynamic characteristics of the portal frame are also identified from the WPT of low frequency signals.

並列關鍵字

Wavelet Transform ； Damage Detection ； System Identification

參考文獻

[1] Sohn, H., Farrar, C.R. Hemez, F.M., Czarnecki, J.J., Stinemates, D.W. and Nadler B.R., 2003, “A review of structural health monitoring literature：1996-2001”, Los Alamos National Laboratory Report, L.A-13976-MS.

[3] Kitada, Y., 1998, “Identification of nonlinear structural dynamic systems using wavelets”, J. Eng, Mech. Div., 124(10), p1059-1066.

[5] Wang, Q. and Deng, X., 1999, “Damage detection with spatial wavelets”, Int. J. Solids Struct., 36(23), p3443-3468.

[6] Hou, Z., Noori, M. and St.Amand, R., 2000, “Wavelet-based approach for structural damage detection”, J. Eng. Mech. Div., Am. Soc. Civ. Eng., 12(7), p677-683.

[7] Coifman, R. R. and Wickerhauser, M. V., 1992, “Entropy-based algorithms for best basis selection”, IEEE Trans. Inf. Theory, 38, p713-718.