This project focuses on the development of methodologies for obtaining reliable Finite Element (FE) models of bridges from measurements of their structural response to ambient as well as earthquake and wind excitations. Comparing such models from different times, for example before and after an earthquake or before and after a maintenance/rehabilitation operation, can provide information on the presence of structural damage and/or on the effectiveness of a rehabilitation solution. These methodologies will be compared with other types of damage detection methodologies that do not require an initial FE model of the structure: these methods try to “recognize a pattern” in the response of the structure corresponding to its healthy state (hence the name of “pattern recognition”). By looking at how this pattern changes, we propose developing a methodology that can assess the presence (amount as well as location) of structural damage.