Extraction of bearing fault features by sliding time-frequency synchronous averaging by maximum amplitudes at potential fault frequencies in envelope spectrum

In early fault diagnosis of bearings, weak fault features tend to be overshadowed by high-energy noise. To address this challenge, this paper proposes a novel algorithm for extracting fault features, termed Sliding Time-Frequency Synchronous Averaging Enhancement based on Maximum Amplitudes at Potential Fault Frequencies in the Envelope Spectrum (STFSA-MAPFFES). In this algorithm, the time-frequency coefficients are innovatively used instead of the time-domain signal for the envelope spectrum calculation. The proposed APFFES parameters utilise the physical properties of the fault to accurately locate the resonance bands in the time-frequency domain. Only the selected time-frequency coefficients are subjected to a sliding time-frequency synchronous averaging process, which achieves efficient early bearing fault feature extraction. Fault features are extracted through two primary outputs: (1) an envelope spectrum that encapsulates the majority of the fault-related information and (2) the time-frequency coefficients, which are further enhanced using the unbiased autocorrelation function and STFSA. A series of digital-analog signals is used to evaluate the performance of the algorithm proposed. Additionally, two publicly available datasets are processed, and the effectiveness of the STFSA-MAPFFES algorithm is compared with six other methods. The results demonstrate that the proposed method outperforms the comparison methods in terms of extracting fault features.