Kaiser window-based S-transform for vibration signal denoising in rolling bearing fault diagnosis

Vibration signal analysis is widely used for bearing fault diagnosis in rotating machinery, but the presence of strong background noise often masks fault-related features, reducing diagnostic accuracy. To address this issue, this paper proposes a noise removal method based on the Kaiser window S-transform (KST), in which the signal is windowed over time using a variable Kaiser window. The Fourier transform is then applied to the windowed signal, yielding a three-dimensional S-transform comprising time, frequency, and transform amplitude. Then, a soft thresholding method is applied to the S-transform for denoising, and finally, the denoised time-domain signal is obtained using the inverse transform. To evaluate the proposed denoising method, it is applied to the experimental bearing vibration signal from a rotating machine. Following denoising, kurtosis is used for fault detection, and the envelope spectrum is used to identify the fault type. Finally, the study compares soft and hard thresholding for noise removal using KST, compares KST with Wavelet Thresholding (WT) under various conditions—including colored noise at different Signal-to-Noise Ratios (SNRs)—and performs fault detection using a Support Vector Machine (SVM) classifier. The results show that the proposed method effectively removes noise while enhancing fault-related features in the signal.