A method is presented for the prediction of the wear history of a train wheel by a numerical simulation. The method consists of three fundamental steps. First, the wheel/rail contact geometry is characterized. Contact points, Herzian contact patch area and creep coefficients are determined as functions of the wheelset lateral position. Second, a time history of the wheelset positions and velocities is determined from a simulation of the vehicle response to a specified track input. Finally, a wear algorithm dependent on the geometry and the dynamics is used to compute the amounts of material to be removed at the discrete points describing the wheel profile. The process is repeated recurrently to simulate the gradual wearing of the wheel. The method has been tested with several different wear models and it has been found that worn profiles are relatively insensitive to the selection of wear model. A parametric study on the effect of creep coefficient and payload on wear rate has been used to differentiate the wear models. Results include predictions of AAR and CN worn wheel profiles. The predictions show that the CN profile wears parallel to the new profile. This trend has been observed in practice. / M.S.