應用小波分析於鋼筋混凝土結構物的損壞識別

Application of Wavelet Analysis to the Damage Detection of RC Frame Structures

陳奕安(I-An Chen)

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(2007年)

https://doi.org/10.6342/NTU.2007.03023

摘要

結構中的破壞位置及損壞程度通常是由有經驗的工程師藉由其肉眼與經驗來檢查及判斷。然而，隨著科技日益進步、人口日漸增多、土地利用日趨密集，結構物也趨向大型化與複雜化，因此結構的行為與潛在的破壞機制變得複雜而不易預測。因此如何有效、可靠地評估結構損壞(包括破壞是否產生、破壞位置及損壞程度)是一件值得探討的課題。

小波分析(wavelet analysis)改善了傅立葉轉換(Fourier transform)只能觀察到頻率域資料的特性，小波分析具有良好的時域(time domain)與頻域(frequency domain)的解析能力。因此本研究利用小波分析的方法，從感知器(sensor)對結構動力反應(dynamic responses)的直接量測訊號中，擷取出一系列的特徵(feature)來識別結構物的損害。小波分析包括了連續小波轉換(CWT)、離散小波轉换(DWT)和小波包轉換(WPT)。本文首先簡短的介紹過去小波分析在結構損壞識別的應用。接著引述小波分析的數學理論，並提出本研究中用來識別結構損害的特徵(feature)，包括保持指數(Holder exponent)、離散小波轉換分解到第一層的細節函數(level-1 details function)、希爾伯振幅譜(Hilbert amplitude spectrum)、希爾伯邊緣譜(Hilbert marginal spectrum)和相位指標(phase index)。

然後先把這些特徵應用到數值模擬的數據去驗證其可靠性。這個模擬模型主要可用來模擬單自由度鋼筋混凝土結構物的動力反應，包括位移反應、速度反應和加速度反應。接著應用到振動台實驗的數據。本研究所使用的振動台實驗數據主要來自鋼筋混凝土構架內填磚牆(RC Brick Wall Frame Structure)面外振動台實驗與標竿鋼構架結構物(Benchmark Steel Frame Structure)振動台實驗。這兩個振動實驗都是在國家地震研究中心執行的。鋼筋混凝土構架是單自由度的結構系統，和前面所提到的數值模擬模型很類似。標竿鋼構架結構物主要想探討此鋼結構物在一樓發生勁度折減所產生的動力反應。研究分析結果顯示：這些特徵對於識別結構破壞的存在、破壞的時間點、破壞嚴重程度、甚至破壞的樓層有不錯的效果。

關鍵字

小波 ； 損壞識別

並列摘要

This study presents a set of feature extraction techniques for damage detection based on wavelet analysis, including continuous wavelet transform, discrete wavelet transform, and wavelet packet transform. Firstly, a brief introduction to wavelet analysis on the structural damage detection is presented. Secondly, a series of description on continuous wavelet transform, discrete wavelet transform, wavelet packet transform, Hilbert transform, and features extracted for damage detection, including Holder exponent, level-1 details function of DWT, Hilbert amplitude spectrum, Hilbert marginal spectrum, and phase index are presented. After that, these features are applied to numerical simulation data to verify their application to damage detection. The numerical simulation data are obtained by a simulation model which was first proposed by Sisvaselvan and Reinhorn (2000), and modified by S. H. Chao and C. H. Loh (2006). The simulation model, named deteriorating hysteretic model, can be used to numerically simulate the dynamic responses of a SDOF RC structural system. After verified by the application to numerical simulation data, these feature extraction techniques are applied to shaking table test data. The test specimens include a RC wall frame structure, and a three-story steel frame structure. The RC wall frame structure can be simplified as a single-degree-of-freedom structural system which is similar to the numerical simulation model mentioned above. The shaking table test of the three-story steel frame structure is studied to identify the abrupt change of stiffness in the first floor of the three-story steel frame structure. Analysis results show that these features, mentioned in this study, can be effectively used to detect structural damage only through the direct response measurement without any a priori knowledge of the original structural system. Based on the analysis results of feature extraction for damage detection, there is a conclusion in the end.

並列關鍵字

wavelet ； damage detection

參考文獻

26. 辜琪媜, “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”, 國立台灣大學土木工程學系碩士論文, 九十四年六月

2. Y. M. Mao, and P. W. Que, (2006) "Application of Hilbert-Huang Signal Processing to Ultrasonic Non-Destructive Testing of Oil Pipelines", J. Zhejiang Univ SCIENCE A, 2006 7(2):130-134

3. Abhijeet Dipak Shinde, (2004) " A Wavelet Packet Based Sifting Process and its Application for Structural Health Monitoring", Master Thesis, August 2004

4. S Park, CB Yun, Y Roh, DJ Inman, (2006) "Active Sensing Network System for Crack Detection on Welded Zone of Steel Truss Member", Proc. of SPIE Vol.6174, 61741A, (2006)

5. J Zhang, RX Li, P Han, DF Wang, XC Yin, (2003) "Wavelet Packet Feature Extraction for Vibration Monitoring and Fault Diagnosis of Turbo-Generator", Proceedings of the Second International Conference on Machine Learning and Cybernetics, Xi'an, 2-5 November 2003