| Damage accumulated in civil and hydraulic structures, such as dams and bridges, could lead to catastrophic collapse of the structure, causing huge loss to the society. Thus, structural health monitoring (SHM) of civil and hydraulic structures is of significant importance. Structural dynamic nondestructive detection (SDNT), as a basic form of SHM, is becoming a focus of research that draws worldwide attention."Structural dynamics + Fourier transform (FT)"is the basic theoretical framework that supports SDNT, while it is still immature due to the inherent deficiency of FT in processing non-stationary damage detection signal. Wavelet transform (WT), on the other hand, overcomes the deficiency of FT and shows potential to drive an advanced"structural dynamics + WT"framework by promoting the"structural dynamics + FT". Nevertheless, relevant research on this aspect is preliminary, and many scientific issues need to be further investigated.This dissertation aims to study five fundamental issues related to promoting the framework of"structural dynamics + FT"to"structural dynamics + WT", based on a combined theoretical analysis, numerical simulation and physical model experiment approach. The following issues are studied:①solving the theoretical detail problems of prevailing intelligent damage detection mode of"wavelet packet transform (WPT) + neural networks (NN)"at present,②developing novel dynamic fingerprints that are immune to the influence of load spectrum,③proposing WT-based stress wave analysis theory that is suitable for stress wave propagation nondestructive detection,④eliminating the deficiencies of current fractal-based damage detection theory, and⑤establishing wavelet-domain self-contained SDNT in physical model experiment. Based on the above critical studies, the major contributions of this dissertation are summarized as follows:(1) Experimental evidence is provided for wavelet packet (WP) sub-band energy as a damage indicator. Both the WP time-frequency and WP time-frequency energy spectrum analysis methods are proposed to create WP sub-band energy damage feature vector of time-frequency property. A method of NN-based principle component analysis is developed to reduce the high dimensionality of WP sub-band energy damage feature vector. Experimental results show that these productions solve the theoretical detail problems involving in the"WPT+NN"intelligent damage detection mode.(2) Dynamic fingerprints of real/complex curvature of frequency response function and real/complex curvature of vibration transmissibility are created, from which the systematic theory of structural transmissibility dynamic fingerprints is formed. In addition, a particular wavelet algorithm is developed for analyzing structural transmissibility dynamic fingerprints of complex-function characteristics. Experimental results show that the dynamic fingerprints are immune to the influence of load spectrum and of good adaptability to engineering practice. The dynamic fingerprints and wavelet algorithm provide theoretical basis for developing viable SDNT techniques.(3) Analytical wavelet-field power density spectrum, real wavelet-field modulus spectrum and modulus maximum spectrum are proposed, and their combination forms a stress wave continuous wavelet-field spectrum analysis strategy. Moreover, a basic approach bringing the strategy into effect is presented based on advantageous combination of analytic Morlet and real Marr wavelet, from which two basic and more practical methods are developed: one dependent of the extended feature of analytic Morlet wavelet and the other on the combination of mended analytic Morlet and improved real Marr wavelet. Compared to the existing WT-based stress wave analysis theory, the proposed strategy and methods are more suitable for stress wave propagation-based nondestructive detection. They have also been successfully applied to build technical routine for identifying damage of foundation medium.(4) A novel SDNT idea is developed to eliminate the deficiencies of current fractal-based damage detection theory. The idea integrates the advantages of both wavelet and fractal to carry out damage detection, i.e.,"to reveal damage information by wavelet and to extract damage feature by fractal". To implement the idea, a special WPT algorithm termed frequency-ordered/time-ordered and non-decimated WPT (FOTO-NWPT) is presented; moreover, a SDNT method combining the FOTO-NWPT and correlation dimension analysis is established. Experimental results show that the FOTO-NWPT enables fractal more effective to extract damage feature than the existing wavelet algorithms, and it eliminates the deficiencies of current fractal-based damage detection theory.(5) An experimental test of utilizing WT to detect the fatigue damage of reinforced concrete beam is conducted. In the experiment, by extending the concepts of frequency response function and vibration transmissibility ratio from frequency domain to wavelet domain, two dynamic fingerprints termed analytic wavelet time-frequency response spectrum (AWTFRS) and analytic wavelet time-frequency transmissibility spectrum are, respectively, established. Furthermore, by exploring the nonlinear abnormalities present in AWTFRS, a self-contained SDNT method is proposed to detect damage by utilizing WT to identify structural nonlinear dynamic characteristics. The fingerprints are critical for developing wavelet-field modal analysis theory, and compared to the general methods, the advanced SDNT method is more applicable for damage detection of practical civil and hydraulic structures due to its capability of surmounting the linear domain to detect structural damage.
|
PDF Full Text Request