Bond–slip behavior at the steel–concrete interface plays a crucial role in the mechanical performance of reinforced concrete and steel–concrete composite structures. Under service conditions, structural loads are inherently random, cyclic, and non-monotonic, making the abrasive wear effect of the steel–concrete interface strongly dependent on the loading path. Conventional cohesive zone models (CZMs) often fail to capture this path-dependent degradation mechanism in a physically consistent manner. In this study, a discrete, path-dependent wear model is proposed within the framework of Alfano’s [1,2] cohesive law. The sliding interface is discretized into multiple segments, each associated with an independent internal wear variable that evolves according to the local sliding history. The interface friction is modeled as a spatially varying field, and the nominal coefficient of friction at each discretized interface segment are directly determined by the associated local wear variable. To ensure numerical stability and physical continuity, a Gaussian smoothing technique is employed to regularize discontinuities in the wear distribution along the interface. The proposed model is capable of reproducing and explaining key observed phenomena in the experiment, including the seating drop effect that is a sudden reduction in tangential force at previously visited positions during cyclic loading, and the ploughing effect that is characterized by an increase in tangential resistance upon first reaching a new sliding position. The resulting traction–separation relationships are systematically analyzed to elucidate the coupled mechanisms of bonding, slipping and abrasive wear. The model provides a robust computational framework for simulating path-dependent interface behavior in composite structures subjected to complex loading history.