Abstract: Previous studies on vehicle-scanning-based bridge damage identification generally rely on baseline comparison and rarely consider the interference of random traffic flow.To address this limitation,this study proposes an iterative damage identification framework under random traffic flow conditions without requiring measured baseline data.First,a sensitivity-analysis-based mapping relationship is established between bridge element damage and the first-mode shape identified from drive-by monitoring,and the residual between the predicted and measured mode shapes is used to formulate the objective function.Next,𝓵_1/2 regularization with sparse representation capability is introduced for damage inversion,and finite element model updating is incorporated to promote convergence of the identification results.By introducing traffic data collected from a weigh-in-motion system installed on an in-service bridge and combining them with Monte Carlo sampling,the model can simulate a random traffic excitation environment representative of practical operating conditions.Numerical examples demonstrate that the proposed method can effectively locate and quantitatively estimate the main damage under both single- and multiple-damage scenarios,although low-amplitude false positives may occur in some local regions.Parametric analyses further indicate that the proposed framework exhibits good robustness to sensor-vehicle speed and traffic flow density,whereas the identification accuracy degrades under high noise levels.Moreover,the proposed framework does not require traffic interruption,thereby taking advantage of the engineering applicability of vehicle scanning methods.