This dissertation studied the structural identification and damage detection of civil engineering structures. Several issues regarding structural health monitoring were addressed. The data-driven subspace identification algorithm was investigated for modal identification of bridges using output-only data. This algorithm was tested through a numerical truss bridge with abrupt damage as well as a real concrete highway bridge with actual measurements. Stabilization diagrams were used to analyze the identified results and determine the modal characteristics. The identification results showed that this identification method is quite effective and accurate. The influence of temperature fluctuation on the frequencies of a highway concrete bridge was investigated using ambient vibration data over a one-year period of a highway bridge under health monitoring. The data were fitted by nonlinear and linear regression models, which were then analyzed. The substructure identification by using an adaptive Kalman filter was investigated by applying numerical studies of a shear building, a frame structure, and a truss structure. The stiffness and damping were identified successfully from limited acceleration responses, while the abrupt damages were identified as well. Wavelet analysis was also proposed for damage detection of substructures, and was shown to be able to approximately locate such damages. Delamination detection of concrete slabs by modal identification from the output-only data was proposed and carried out through numerical studies and experimental modal testing. It was concluded that the changes in modal characteristics can indicate the presence and severity of delamination. Finite element models of concrete decks with different delamination sizes and locations were established and proven to be reasonable. Pounding identification can provide useful early warning information regarding the potential damage of structures. This thesis proposed to use wavelet scalograms of dynamic response to identify the occurrence of pounding. Its applications in a numerical example as well as shaking table tests of a bridge showed that the scalograms can detect the occurrence of pounding very well. These studies are very useful for vibration-based structural health monitoring.