Wheel sliding frequently occurs under poor wheel-rail adhesion status in railroad vehicles, which may cause abnormal abrasions of the wheel and rail interface. Identifying the wheel-rail adhesion status can contribute to precise anti-slip control and improving adhesion utilization; correspondingly, the abnormal abrasions can be reduced. An approach for identifying wheel-rail adhesion status based on an improved recursive Levenberg-Marquardt (RLM) algorithm is proposed, and the Polach model for tangential force calculation is used to fit the nonlinear creep force. The local gradient of objective function is modified by Hessian matrix to realize the adaptive change of the algorithm. Moreover, a variable forgetting factor based on the sigmoid function is introduced to improve to the robustness and response speed of identification algorithm. Simulation results indicate that the proposed algorithm can accurately identify the time-varying wheel-rail adhesion status under large creepage conditions.