This thesis investigates the system identification of the Inverso Pinasca bridge, a structure that has been affected by natural phenomena, including water-induced scour, which have compromised the stability of its piers. The principal objective is to identify the bridge's modal parameters, which are crucial for assessing its structural integrity and identifying requisite interventions. By making use of advanced structural health monitoring (SHM) techniques, the study is concerned with the robustness of machine learning-based automated operational modal analysis (AOMA) algorithms in the presence of both stationary and non-stationary signals. A key challenge inherent to traditional system identification algorithms is their dependence on stationarity assumptions and the absence of explicit guidelines concerning signal duration.