A multi-synchrosqueezing ridge extraction transform for the analysis of non-stationary multi-component signals

Abstract

Condition monitoring (CM) signals of rotating machines operating under varying speed condition typically exhibit amplitude modulation and frequency modulation characteristics. A recent study [G. Yu, T. R. Lin. Mech. Syst. Signal Process. 147 (2020) 107069] shows that multi-synchrosqueezing transform (MSST) can effectively extract the distinctive time frequency features from non-stationary signals using an iteration process in conjunction with the synchrosqueezing transform. However, the noise contained in a signal can become a serious problem as the number of iterations increases in the transform. An alternative time-frequency analysis (TFA) method blending a ridge extraction technique and a MSST transform is thus proposed in this study to overcome the noise interference problem. In this approach, the ridge extraction technique is used to extract each mono component contained in the TFA results of the MSST in turn. A noise-free time frequency representation can then be reconstructed by superimposing the time frequency distributions of all mono-components for an accurate fault diagnosis of rotating machines under varying speed condition. A peak-hold-down-sample (PHDS) algorithm is also utilized in this work to improve the computation efficiency and to avoid possible computer jamming caused by large data. electronic document is a “live” template.