In the dissertation, the vibration energy distribution in structural acceleration responses is investigated, and a novel two-step damage identification procedure based on this quantity is introduced. Many existing damage identification methods rely on a mechanical model of the considered structure. The advantage of the proposed procedure is that it avoids this restriction as far as possible. In the first step damage detection: Is there a damage? no modal parameters are required, because it is completely based on the vibration energy / vibration characteristics of response measurements. The normalized cumulative power spectral density (NCPSD) is defined as the damage sensitive feature. Its dissimilarity in an undamaged reference state and in the actual structural state is expressed by a so-called NCPSD damage index. This flexible quantity can be computed for individual structural nodes, averaged over the whole structure, or also statistically evaluated in long-term monitoring, since the evaluation is completely automatable. Variation of temperature and excitation can be taken into account. Only for the second step damage localization: Where is the damage? it was found that identified mode shape data is necessary for successfully locating the damage, based on two defined mode shape damage indexes. The proposed two-step damage identification procedure is tested on several models simulated on the computer and on three real structures. It is shown that the NCPSD damage index is suitable to indicate most of the imposed structural modifications, representing damage, despite the unknown nature of the ambient excitation. In a simulated long-term monitoring project, the statistical characteristics of the NCPSD damage index indicate damage successfully. The utilized mode shape damage indexes are shown to support the localization of the spot or domain of modification/damage.