Abstract:  Aiming at the problem that the strong vibration and noise of rolling bearings were usually superimposed with the early weak faults of the bearing, which made it difficult to extract the transient fault characteristics, a time-frequency analysis method was proposed to analyze the transient characteristics of the early weak fault signals of the bearing. Firstly, the data was preprocessed by complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), the effective modal components were filtered out using kurtosis and were reconstructed to reduce noise. Then,the reconstructed signal was subjected to transient extraction transform (TET) for transient feature extraction. Finally, the fault diagnosis was performed using the extracted transient signals. The simulation signal and experimental signal were processed and compared with other common time-frequency analysis methods. The results show that this method can effectively extract the fault transient characteristics, improve the noise robustness of early fault signals of rolling bearing under complex environment, focus the time-frequency energy characteristics more clearly, clearly see the interval of transient signals, and can effectively represent the early fault characteristic frequency of the signal.

Key words:  rolling bearing; complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN); time-frequency analysis(TFA); effective modal components;transient-extracting transform (TET)