Time-extracting S-transform algorithm and its application in rolling bearing fault diagnosis

• Published in: Science China. Technological sciences Vol. 65; no. 4; pp. 932 - 942
• Main Authors: Xu, YongGang, Wang, Liang, Hu, AiJun, Yu, Gang
• Format: Journal Article
• Language: English
• Published: Beijing Science China Press 01.04.2022; Springer Nature B.V
• Subjects: Algorithms; Engineering; Fault diagnosis; Fourier transforms; Parameter uncertainty; Roller bearings; Signal processing; Time domain analysis; Time-frequency analysis; time-extracting S-transform; fault diagnosis; pulse signal; short-time Fourier transform; time-frequency analysis
• ISSN: 1674-7321; 1869-1900
• DOI: 10.1007/s11431-021-1919-y

Time-frequency (TF) analysis (TFA) is one of the effective methods to deal with non-stationary signals. Due to their advantages, many experts and scholars have recently developed post-processing algorithms based on traditional TFA. Among them, short-time Fourier transform (STFT) based post-processing algorithms have developed the fastest. However, these methods rely heavily on the window length selected in STFT, which has great influence on the post-processing algorithm. In this paper, a post-processing algorithm for effectively processing pulse signals was proposed and called time-extracting S-transform (TEST). The time-domain extraction method based on S-transform avoids the influence of uncertain parameters. After comparing the performance of various TFA methods when processing analog signals, the proposed TEST can clearly show the pulse occurrence time under the premise of ensuring high TF aggregation. The actual signal proves that the method can be used for fault diagnosis of rolling bearings.