The wheel-rail vibration signal of metro vehicle has the following characteristics: low signal to noise ratio, nonlinear, nonstationary and so on. To achieve better extraction to the fault feature of flat wheel, a fault diagnosis method of metro vehicle flat wheel based on envelop spectrum entropy and Variational Mode Decomposition (VMD) was proposed. Firstly, the simulation signal was used for VMD analysis, comparing with empirical mode decomposition, to illustrate VMD , s effectiveness. Then, the four signals from different condition obtained from experiments were applied for the VMD, to calculate the envelop spectrum entropy of different decomposition modes. Finally, the effect of fault diagnosis was analyzed by support vector machine. The experimental result showed that the proposed method in this paper could effectively achieve the flat wheel fault diagnosis of metro vehicles, which had a good effect on the fault diagnosis of the metro vehicle.