Compound fault diagnosis method of rolling bearings based on multi-parameter adaptive feature mode decomposition

ObjectiveCoupled composite faults, unbalanced fault feature amplitudes, and fault impacts easily masked by strong background noise commonly exist in rolling bearings. A rolling bearing composite fault diagnosis method based on multi-parameter adaptive feature mode decomposition was proposed to address the above problems.MethodsFirstly, the golden jackal optimization algorithm was improved by multiple strategies, and the compound periodic harmonic energy ratio fitness function and parameter assignment exchange strategy were introduced to realize adaptive optimal setting of constrained parameters, including mode number, filter length and filter number. Secondly, multi-mode decomposition of vibration signals was implemented by the multi-parameter adaptive feature mode decomposition algorithm. Then, sensitive mode components were screened according to the maximum periodic harmonic energy ratio index. Finally, envelope spectrum analysis was performed on sensitive components to identify fault features, and quantitative comparison with typical diagnosis methods was carried out.ResultsThe results show that the proposed method can effectively separate composite fault features of bearing inner and outer races. Fault frequencies and their modulation sidebands can be clearly identified under the strong noise condition with a signal-to-noise ratio of -12 dB. Compared with feature mode decomposition and improved feature mode decomposition, the proposed method yields higher kurtosis and squared envelope Gini index, and exhibits better performance in fault decoupling separation and noise suppression.