A post-processing method called Fourier transform based on local maxima of autocorrelation function for extracting fault feature of bearings

Abstract

Accurate extraction of fault features from signals containing weak repetitive pulses is a critical issue in fault diagnosis of bearings. To address this challenge, a novel solution termed Fourier transform based on Local Maxima (FT-LM) is proposed in this paper. In the FT-LM method, local maxima are first extracted from signals, followed by the application of a Fourier transform to this sequence. Furthermore, to enhance the detection of repetitive pulses, the Short-Time Fourier Transform (STFT) and the unbiased autocorrelation function are integrated into the FT-LM, developing a post-processing algorithm called Fourier Transform based on Local Maxima of Autocorrelation Function (FT-LMACF). By employing a series of simulated signals and two sets of public experimental data, comparisons are made among the FT-LMACF, mainstream post-processing methods, the Kurtogram method and the spectral coherence based on the STFT. The results indicate that the FT-LMACF outperforms other algorithms in extracting fault features inherent in repetitive pulses.