Detection of structural damage in civil engineering, particularly in critical infrastructures such as bridges, is a major research concern on a global scale. This research aims to present a non-invasive methodology employing the Proper Orthogonal Decomposition (POD) technique to identify unstable locations within bridge structures. Data for this investigation were collected from wireless sensors deployed on an operational highway bridge in Upper State New York. The offline data used for POD analysis consisted of acceleration data from two distinct scenarios called “Baseline configuration” and “Damage Scenario 1”. The analysis involved the reconstruction of the measured acceleration data using the most dominant POD basis functions. Subsequently, predictions for the acceleration data for a scenario called “Damage Scenario 2” were calculated. In order to identify the location of the damage, a statistical measure called kurtosis was applied. The results reveal that accelerometers near the first diaphragm exhibited higher kurtosis values compared to those located farther away. The results of this study demonstrate a strong correlation between the predictions and the measured data, providing an effective means to identify areas of instability within the damaged bridge. This research contributes to the advancement of non-invasive damage detection techniques, with significant implications for civil engineering studies.