A novel algorithm for extracting bearing fault features based on cyclostationary and impulsive characteristics

Extracting distinct fault features is essential for monitoring bearing health; however, it becomes particularly challenging when fault components are weak. To address this issue, an enhancement algorithm for bearing fault features based on cyclostationary and impulsive characteristics is proposed. The algorithm systematically integrates unbiased autocorrelation function, continuous wavelet transform, time marginal spectrum, and sliding time synchronous averaging to progressively enhance fault features in both time and frequency domains. A key innovation of the algorithm lies in its ability to leverage physical characteristics of fault pulses to effectively suppress noise and amplify fault components. The experimental validation demonstrates the algorithm’s effectiveness in extracting fault features under low signal-to-noise ratio conditions. Comparative analysis with benchmark methods, including Fast Kurtogram, Autogram, spectral correlation based on short-time Fourier transform, and Adaptive Maximum Second-Order Cyclostationary Blind Deconvolution, confirms its superior performance in isolating fault pulses and prominently highlighting fault frequencies in envelope spectrum. By increasing probabilities of feature detection, and extending lead times for residual‑life estimation, potential faults are advanced from a latent phase to an observable phase, significantly improving timeliness of predictive maintenance and inherent operational safety of bearings.