Concave sparsity-assisted time-reassigned multisynchrosqueezing transform for early-stage weak fault feature enhancement with applications to bearing fault diagnosis

Early-stage defects in wind turbine bearings often generate weak impulsive signatures that are easily obscured by strong background noise, posing a significant challenge to accurate extraction of fault-related features. To overcome this limitation, a novel concave sparsity-assisted time-reassigned multisynchrosqueezing transform (CSA-TMSST) is developed to improve the extraction of weak fault-related signatures. First, a non-convex sparse optimization model incorporating the generalized minimax-concave (GMC) penalty is established to reconstruct fault impulsive components from the raw signal. The introduced GMC-based sparse regularization enhances signal sparsity, enabling the accurate extraction of impulsive fault signatures. Subsequently, TMSST is employed to characterize the reconstructed signal in the time-frequency (TF) domain, yielding a more concentrated time-frequency representation (TFR). Furthermore, the iterative reassignment mechanism is utilized to enhance the concentration of TF energy distribution, thereby yielding a clearer TFR of transient fault features. Finally, characteristic fault frequencies are identified via spectral analysis. The results demonstrate that CSA-TMSST is capable of extracting weak fault-related signatures and producing high-quality TFRs under strong noise conditions, confirming its effectiveness and superiority.