This thesis addresses the experimental and numerical study of vibration-based damage detection (VBDD) techniques in structural health monitoring (SHM) of bridge superstructures. The primary goal of SHM is to ascertain the condition or health of a structure so that decisions can be made with regard to the need for remediation. VBDD techniques are global non-destructive evaluation (NDE) techniques. The principle of VBDD techniques is to detect damage using changes in the dynamic characteristics of a structure caused by the damage. The advantage of VBDD techniques over local NDE techniques is that VBDD techniques can assess the condition of an entire structure at once and are not limited to accessible components. <p>Well controlled laboratory experiments on a half-scale, simply supported steel-free bridge deck and two full-scale, simply supported prestressed concrete girders demonstrated that small scale damage at different locations can be reliably detected and located by VBDD techniques using a relatively small number of sensors (accelerometers or strain gauges) and considering changes to only the fundamental mode of vibration. The resolution of damage localization, defined as the length of the window within which damage could be located when the technique predicts it to be located at a particular point, was 70% of measurement point spacing for the deck and 82% for the girders, provided the damage was not located too close to a simple support.<p>To establish the potential of VBDD techniques in the absence of experimental uncertainty, eigenvalue analyses using finite element models of the deck and the girders were undertaken to investigate ability of five VBDD methods to predict the longitudinal location of damage. It was found that when mode shapes were well-defined with a large number of measurement points, the damage location could be determined with great accuracy using any of the five VBDD techniques investigated. The resolution of longitudinal localization of damage was 40 to 80% of the spacing between measurement points when small numbers of measurement points were used, provided the damage was not located too close to a simple support.<p>The experimental study successfully detected small scale damage under forced resonant harmonic excitation but failed in detecting damage under forced random excitation, although the use of random sources of excitation is more practical in field testing. Transient dynamic analyses on the finite element model of the steel-free bridge deck were performed to investigate the implications of using random forced vibrations to characterize mode shapes to be used to detect damage. It was found that the probability of successful damage localization depends upon the severity of the damage, the number of trials used to obtain the average mode shape, the location of damage relative to the nearest sensor, the distance between the damage and the support, and the magnitude of measurement errors. A method based on the repeatability of measured mode shapes is proposed to calculate the probability of successful damage detection and localization.<p>In summary, results of this research demonstrate that VBDD techniques are a promising tool for structural health monitoring of bridge superstructures. However, although these methods have been shown to be capable of effectively detecting small scale damage under well controlled conditions, a significant amount of challenging work remains to be done before they can be applied to real structures.

This thesis addresses the experimental and numerical study of vibration-based damage detection (VBDD) techniques in structural health monitoring (SHM) of bridge superstructures. The primary goal of SHM is to ascertain the condition or health of a structure so that decisions can be made with regard to the need for remediation. VBDD techniques are global non-destructive evaluation (NDE) techniques. The principle of VBDD techniques is to detect damage using changes in the dynamic characteristics of a structure caused by the damage. The advantage of VBDD techniques over local NDE techniques is that VBDD techniques can assess the condition of an entire structure at once and are not limited to accessible components. <p>Well controlled laboratory experiments on a half-scale, simply supported steel-free bridge deck and two full-scale, simply supported prestressed concrete girders demonstrated that small scale damage at different locations can be reliably detected and located by VBDD techniques using a relatively small number of sensors (accelerometers or strain gauges) and considering changes to only the fundamental mode of vibration. The resolution of damage localization, defined as the length of the window within which damage could be located when the technique predicts it to be located at a particular point, was 70% of measurement point spacing for the deck and 82% for the girders, provided the damage was not located too close to a simple support.<p>To establish the potential of VBDD techniques in the absence of experimental uncertainty, eigenvalue analyses using finite element models of the deck and the girders were undertaken to investigate ability of five VBDD methods to predict the longitudinal location of damage. It was found that when mode shapes were well-defined with a large number of measurement points, the damage location could be determined with great accuracy using any of the five VBDD techniques investigated. The resolution of longitudinal localization of damage was 40 to 80% of the spacing between measurement points when small numbers of measurement points were used, provided the damage was not located too close to a simple support.<p>The experimental study successfully detected small scale damage under forced resonant harmonic excitation but failed in detecting damage under forced random excitation, although the use of random sources of excitation is more practical in field testing. Transient dynamic analyses on the finite element model of the steel-free bridge deck were performed to investigate the implications of using random forced vibrations to characterize mode shapes to be used to detect damage. It was found that the probability of successful damage localization depends upon the severity of the damage, the number of trials used to obtain the average mode shape, the location of damage relative to the nearest sensor, the distance between the damage and the support, and the magnitude of measurement errors. A method based on the repeatability of measured mode shapes is proposed to calculate the probability of successful damage detection and localization.<p>In summary, results of this research demonstrate that VBDD techniques are a promising tool for structural health monitoring of bridge superstructures. However, although these methods have been shown to be capable of effectively detecting small scale damage under well controlled conditions, a significant amount of challenging work remains to be done before they can be applied to real structures.