ISSAFMD: an improved spectral subtraction assisted feature mode decomposition method for acoustic fault diagnosis of flexible thin-walled bearing

Abstract

The acoustic fault diagnosis (AFD) technology, capable of extracting fault features from acoustic signal obtained through noncontact measurements, provides an effective solution to the condition monitoring of flexible thin-walled bearing (FTWB) in dynamic running conditions. However, the acoustic signal is severely affected by environmental noise and coupling noise generated by cascading vibrations between multiple components, which weaken the FTWB’s fault features and lead to the failure of AFD. Therefore, this article proposes a two-stage AFD method called improved spectral subtraction (ISS)-assisted feature mode decomposition (ISSAFMD), to contrapuntally suppress environmental noise and coupling noise. Specifically, an ISS method, with an adjustable SS factor and a power spectral exponent, is proposed to regulate the degree of environmental noise suppression. Additionally, adjustable width Gaussian windows are introduced to replace Hanning windows used in FMD, aiming to improve the decoupling accuracy and efficiency of FMD for multiple components. The analysis results of the simulated signal and experiment signal demonstrate that the proposed ISSAFMD method can effectively suppress the environmental noise and separate weak fault signatures in the AFD of FTWB. Moreover, it outperforms some the state-of-art signal processing methods.