Abstract: Based on global transmissibility functions and principal component analysis, eigenvalues are extracted to construct damage indices. These indices are then combined with the statistical properties to establish detection thresholds, proposing a novel method for effective structural damage identification. Unlike conventional approaches, this method eliminates the need for pre-selecting global transmissibility and replaces frequency band selection with principal components’ selection, reducing reliance on subjective judgment and improving damage detection accuracy. Numerical simulations and experimental tests demonstrated effectiveness of the proposed method on both single and multiple damage scenarios across diverse structural forms, including continuous beams, spatial steel trusses, and steel signal towers under varying loads. Even under high noise levels or minor localized damage, while the damage localizations may extend to adjacent regions, the method still provide reasonable and reliable results.