Local Optimum Time-Reassigned Synchrosqueezing Transform for Bearing Fault Diagnosis of Rotating Equipment

The impulse features are typical characteristics in vibration signals when rotating equipment failures occur, such as faults of bearings and gears. Accurate capture of impulse features can help better monitor the operating status of equipment. As an energy-concentrated time-frequency analysis (TFA) tool, time-reassigned synchrosqueezing transform (TSST) is an effective means to characterize impulse features in vibration signals. However, the existence of some obvious defects limits its application. For example, strong frequency-varying signals cannot be effectively processed, and in a noisy environment, transient impact characteristics cannot be accurately captured. To improve the performance of TSST, we use the properties of the window function to search for the local optimum of the short-time Fourier transform (STFT) result to obtain a new time rearrangement operator termed as local optimum time rearrangement operator. Compared with the original rearrangement operator, this operator can more accurately capture the instant when the transient impact occurs when addressing strong frequency-varying signals. With the assistance of this operator, an effective TFA technique termed local optimum TSST (LOTSST) can be constructed. Compared with TSST, LOTSST can generate time-frequency representation (TFR) with more concentrated energy and has better noise robustness. In addition, LOTSST allows the transient characteristics to be completely reconstructed, from which we can more intuitively observe the instant when the impact occurs. Finally, this article employs the proposed method for bearing fault diagnosis, considering both constant speed and variable speed conditions. The simulation and experimental results show that the proposed method is promising and competitive with other TFA technologies.