Abstract Civil engineering structures are generally supposed to have long operation life spans during which they are submitted to many factors, both environmentally and human-induced, that degrade their condition and performance, gradually leading to the occurrence of damages and malfunctions. Therefore, systems capable of monitoring and assessing structural performance automatically are highly desirable. In this context, this paper describes the vibration-based structural health monitoring of Baixo Sabor arch dam, in Portugal, detailing results from the several processing stages needed to achieve the ability to detect damage. Automated operational modal analysis is applied to the data obtained during the first three years of monitoring, after which environmental and operational conditions affecting modal properties are studied and their effects minimised through the combined application of weighted regression models and principal components analysis. Finally, damages are simulated in a numerical model of the dam, and the results are used to test the ability of damage detection tools to find these anomalies. The present paper uses a quite unique set of experimental data to test for the first time the ability of a vibration-based SHM system to identify realistic damages in an arch dam.