It is usually indicated a defect in the rotating machine by the impulse features in the condition monitoring (CM)signals. To accurately capture the impulse components in the CM signal, this paper proposes a high-precision time-reassigned multi-synchrosqueezing W transform (TMSSWT), which improves the time-frequency concentration performance of the CM signal. The algorithm first uses the W transform （WT） to obtain the time-frequency representation result of the signal, and then constructs a time-frequency post-processing operator - “synchrosqueezing” operator to estimate the group delay（GD）of the signal in the time-frequency domain, thus performing a "squeeze" operation on the time-spectrum energy obtained from the original WT. Secondly, a fixed-point iterative algorithm is used to reallocate the time-frequency energy to the real GD ridge of the signal. Finally, we employ the impulse feature extraction algorithm of signal reconstruction to calculate the time-frequency envelope and then extract the impulse characteristics of the frequency corresponding to the strongest amplitude from the time-frequency spectrum obtained by TMSSWT. By comparing the time interval of impulse characteristics, the bearing fault can be identified. It is well proved that the proposed method can effectively improve the time-frequency energy concentration through the results of simulated signal, in order to accurately capture the impulse characteristics in the actual fault diagnosis application, and better identify and diagnose the mechanical fault bearing.