High-order Synchrosqueezing Superlets Transform and its application to mechanical fault diagnosis

Abstract

Under the condition of variable speed, the vibration signal of mechanical transmission components present the typical non-stationary characteristics. Nevertheless, some problems, such as insufficient time–frequency energy concentration and frequent noise disturbance, exist in the current signal processing methods during the extraction of time-varying fault characteristics. Different from the frequently-used time–frequency analysis (TFA) methods such as Continuous Wavelet Transform (CWT) and Synchrosqueezing Transform (SST), High-order Synchrosqueezing Superlets Transform (HSSLT) is proposed in this paper in order to obtain a clearer and more centralized time–frequency representation (TFR). Firstly, multiple groups of wavelets with increased bandwidth are combined into a super wavelet set, and then the Superlets Transform (SLT) is proposed. The single wavelet mother function in the traditional CWT is replaced by this superlets set, which has better wavelet diversity and analytical ability for complex multi-component signals. On this basis, in order to further improve the accuracy of instantaneous frequency (IF) estimation and the energy concentration of TFA, the superlets coefficients calculated by SLT are used for the high-order IF estimation and time–frequency energy rearrangement, thereby HSSLT is presented in this paper. It has obvious advantages of high-resolution TFR and strong noise robustness. Finally, through the rolling bearing fault diagnosis of test bench and the 850 kW offshore wind turbine, the good application potential of the proposed HSSLT in the mechanical fault diagnosis under variable speed conditions is verified.