Synchro-Reassigned Extracting Transform and Its Application to Bearing Fault Diagnosis under Variable Speed Condition

High-resolution time-frequency representation is critical for signal analysis and condition monitoring. synchroextracting transform based on frequency or time reassignment are new types of nonstationary signal processing methods, and their performance is better than that of conventional methods when analyzing time-varying signals. However, the limitation is that they cannot accurately analyze signals that contain both “slowly-varying” and “rapidly-varying” features. To avoid the disadvantages of SET, this paper proposes a novel strategy called Synchro-Reassigned Extracting Transform (SRET) to process nonstationary signals with different modulation characteristics. By using the instantaneous frequency operator and the group delay operator, SRET reassigns and extracts the time-frequency coefficients synchronously in the frequency and time directions to achieve sharpening of energy ridges. To use the computer for fast calculation, the paper also provides a discretization implementation algorithm. Finally, the proposed approach has been applied to numerical simulations and application research. The results show that SRET can accurately estimate the time-varying characteristics of nonstationary signals, and has the potential for fault diagnosis of rotating machinery.