Nonstationary Fault Feature Extraction of Rotating Machinery Using the Synchrosqueezed Wavelet Transform

Quick response to fault occurrence of rotating machinery relies mainly on an accurate detection for the change of frequency constitution in vibration signals. As modern rotating machinery becomes more complex with an increasingly extreme working condition, the feature characterization of nonlinear and nonstationary data can be easily smeared by traditional time-frequency (TF) analysis tools. The Hilbert-Huang transform (HHT), being well-known for its data adaptiveness and its resulted narrow-band intrinsic mode functions (IMFs), suffer 'mode mixing' problem when characterizing multi-component signals in practical applications. In this contribution, the synchrosqueezed wavelet transform (SWT), is facilitated to characterize the multi-component signal by decoupling the harmonic and impulsive components in the simulated signal. The case studies further conducted on a run-up misaligned rotor kit and a surge fault centrifugal compressor justify our conclusions. The corresponding energy distributions achieve a sharper positioning accuracy simultaneously along the time- and frequency-axis.