1.1 Background
Structural system identification and damage detection have received more and more attention in the field of civil engineering. Through monitoring data on structures, a quantity of information can be obtained. A well-known classification for damage identification methods can be defined as four levels: (Level-1) Determination or detection that damage is present in the structure; (Level-2) Determination of the geometric location of the damage; (Level-3) Quantification of the severity of the damage; (Level-4) Prediction of the remaining service life of the structure. One of the efficient and accurate damage detection techniques applicate to all types of structural systems is the vibration-based damage detection (VBDD). The vibration characteristics of a structure can be considered to be a global response signature that can be used as the basis for assessing its condition because they contain embedded information about the structure’s inherent properties. Changes in the structural condition will be reflected in the vibration signature, which make it possible to identify the presence of damage by tracking changes to that signature. Fourier-based analysis has been used as a means of translating vibration signals from time domain into the frequency domain to detect the vibration signatures. But the Fourier transform is not able to present the time dependency of signals and it cannot
measured from the vibration of structures. Wavelet transform can be viewed as an extension of the traditional Fourier transform with the adjustable window location and size, which has recently emerged as a promising tool for structural health monitoring (SHM) and damage detection due to its inherent properties. Its widespread use is also due to its availability of fast and accurate computational algorithms for signal transformation and reconstruction.
Early work using wavelet analysis for SHM has been carried out from several perspectives. From a system identification perspective, Basu and Gupta (1997) applied wavelet analysis to obtain the spectral moments and peak structural responses of multi-degree-of-freedom (MDOF) systems subjected to nonstationary seismic excitations.
Staszewski (1997) used time-scale decomposition to identify the damping in MDOF systems. Todorovska (2001) used the continuous wavelet transform to estimate the instantaneous frequency of signals. Kijewski and Kareem (2003) used wavelet analysis for system identification. Lardies and Ta (2005) used a wavelet-based approach to estimate the instantaneous frequency, damping, and envelope of the system. Li and Liang (2012) proposed a generalized synchrosqueezing transform to enhance the signal time-frequency representation. Tarinejad and Damadipour (2014) used modified Morlet wavelet to estimate damping. Klepka and Uhl (2014) identified the modal parameters of non-stationary systems with the recursive method based on the wavelet adaptive filter.
Chen et al (2014) detected the sudden stiffness reduction in acceleration time history using discrete wavelet transform. Guo and Kareem (2015) utilized the transformed singular value decomposition in tandem to automate the identification of analysis regions in the time-frequency domain. Subsequently, Laplace wavelet filtering is adopted to extract impulse-type signals from the WT coefficients to estimate the damping from transient nonstationary data. From a signal processing perspective, Robertson et al (2003) used Holder exponent based on the wavelet transform to detect the presence of damage and determine when the damage occurred. Goggins et al (2007) divided wavelet coefficients into several frequency bands and the degree of correlation between coefficients of ground and response acceleration was evaluated and allowed yielding and buckling events to be detected. Nair and Kiremidjian (2009) use a wavelet energy based approach to detect the damage. Noh et al (2011) use wavelet-based damage-sensitive features (DSFs) extracted from structural responses recorded during earthquakes to diagnose structural damage. Lee et al (2014) proposed a continuous relative wavelet entropy-based reference-free damage detection algorithm for truss bridge structures and showed that it was sensitive to slight damage extent for the tested damage type (i.e.
loosening of bolts). Balafas and Kiremidjian (2015) developed and validated a novel earthquake damage estimation scheme based on the continuous wavelet transform of input and output acceleration measurements. Amezquita-Sanchez and Adeli (2015) used
the synchrosqueezed wavelet transform fractality model to detect, locate, and quantify the damage in smart high-rise building structures.
1.2 Research objectives
The objective of this study is to utilize the pattern-level feature extraction technique through the continuous wavelet transform (CWT) with the proposed modified complex Morlet wavelet with variable central frequency (MCMW+VCF) to decompose the vibration responses into wavelet coefficient distribution as a joint function of time and frequency. The flowchart of the pattern-level feature extraction technique is shown in Figure 1-1. Different from centralized feature extraction, pattern-level feature extraction technique uses response measurement from individual sensing node. Then based on the extracted features, several damage assessment techniques can be implemented using the pattern-level fusion techniques.
The organization of this study is briefly described as follows:
Chapter 1: A brief description of the research background and the literature survey on the existing identification techniques based on the wavelet transform is presented. Then a general introduce to the objective and scope of this research.
Chapter 2: Introduce the wavelet analysis including the continuous wavelet transform (CWT), discrete wavelet transform (DWT), and wavelet packet transform (WPT). More
details of the CWT will also be described. By making use of the CWT, vibration features of structure can be extracted and then applied to structural damage identification.
Chapter 3: In this chapter, 5 damage assessment algorithms based on the extracted features will be described, which include: (1) Marginal spectrum, (2) Central frequency, (3) Novelty index, (4) Correlation and (5) Normalized component energy. Response measurement from two lab experiments will be firstly introduced then used to validate the proposed methods.
Chapter 4: Based on the features extracted from earthquake response data, 4 algorithms can be used to identify the damage, which include: (1) Marginal spectrum, (2) Central frequency, (3) Pseudo-instantaneous frequency, and (4) Unwrapped phase. Also, the same two experiments from shaking table tests that have been introduced in chapter 3 will be utilized to verify the algorithms.
Chapter. 5: The earthquake response data collected from two building structures under earthquake excitation will be used to verify the proposed methods.
Chapter. 6: Discussion and conclusion for the use of the proposed methods will be given.
The future work of this topic will also be indicated.