Parameterized resampling multisynchrosqueezing transform for the analysis of bearing non-stationary signals

Abstract The vibration signals of bearings operating at time-varying speeds exhibit remarkably non-stationary and typically are accompanied by noise interference. The time-frequency distribution (TFD) obtained by the traditional time-frequency analysis (TFA) methods is blurred due to the resolution limitation. For accurately revealing instantaneous fault characteristic frequencies of signals, parameterized resampling multisynchrosqueezing transform is put forward, which combines the advantages of the parameterized resampling time-frequency (PRTF) transform and the improved multisynchrosqueezing transform. Firstly, instantaneous frequency components are extracted by means of the PRTF transform and its inverse transform to reduce noise interference. The time-frequency energy concentration is subsequently improved by the resampling operator and reassignment operation. Finally, the simulation signal and two real-world bearing fault signals are applied for validation. The results indicate that the TFD obtained by the proposed method exhibits the highest energy concentration and the best readability under noise interference, as well as excellent anti-noise robustness, in comparison with other TFA methods.