This paper proposes a wavelet decomposing technique using the average of the vertical vibration signatures generated at bogie frame with a view to developing the technique as an on-board rail wheel flat monitoring system. Although vibration signatures are affected by a large number of design and operational parameters, such as the track irregularities, the train bogie system damping and variation in loading pattern, the proposed method aims at overcoming these difficulties to provide clear insight into the identification of the existence of wheel flats even at higher operating speeds. The wavelet decomposition method presented in this paper shows that the absolute values of the 4th level detail component of the acceleration signature averages clearly exhibit the existence and location of the wheel flat damages. One of the advantages of the method is its ability to display the 4th level detail component as a 2D plot that is easy to interpret rather than the 3D plots traditionally used by other researchers as presented in the literature. To illustrate this approach, vertical acceleration signatures of a railway containing flat wheels in the front axles have been generated using a 2D Matlab-Simulink dynamic simulation system model previously developed by the authors. The simulation results obtained for different operational speeds on a track containing large irregularities presented in this paper, demonstrate that the wavelet decomposition model is an efficient and easy method to identify wheel flats such that it could be integrated into whole-of-train on-board real-time condition monitoring systems.