Under the interference of noise signals, the vibration impact induced by the defect of aviation bearing is too small to be directly detected from the time domain and frequency domain, especially for the incipient defect. Moreover, there are some shortcomings in existing methods for the accurate diagnosis of bearing defects. To overcome this problem, an improved short-time dimensionless time domain statistical value (TDSV) calculation method to monitor the quality state of bearings during the whole life cycle has been proposed in this paper. The root-mean-square (RMS) of kurtosis is used to describe the total energy of the defect impact features. Based on the bearing data of the Case Western Reserve University (CWRU), the result shows that the kurtosis and its RMS are the most suitable values for bearing quality monitoring compared with other TDSVs. When data interception length is in a certain range, the kurtosis can accurately reflect the energy of the transient vibration of the bearing. The RMS of kurtosis can reflect the defect state of the bearing. By comparing with the experiment data of aviation bearings, the result shows that the kurtosis and its RMS can be applied in the quality monitoring of aviation bearings.