Adaptive Multisynchrosqueezing Transform for Vibration Signal Analysis of Rotating Machinery

As a typical nonstationary signal, the strong frequency modulation (FM) characteristics of rotating machinery vibration signals pose a significant challenge to traditional time-frequency analysis (TFA) methods. Conventional TFA methods are unable to obtain a clear time-frequency (TF) spectrogram when processing strong FM signals, and often suffer from energy diffusion. Post-processing methods can improve energy concentration to some extent. However, most post-processing methods are based on the short-time Fourier transform, which inevitably causes amplitude distortion because of the constant window lengths when facing strong FM signals. To address this issue, a new TFA method is proposed, which is known as adaptive multisynchrosqueezing transform (AMSST). The window scale of AMSST can be adaptively adjusted depending on the slope of instantaneous frequency (IF) ridges to match the non-stationary characteristics. Furthermore, the frequency-reassignment operator is improved on the basis of the variable window. The simulation results demonstrate that the method exhibits excellent energy concentration and noise robustness. The method is subsequently applied to actual bearing fault signals and wind power planetary gearbox fault signals in order to validate the method’s efficacy in the analysis of mechanical vibration signals.