A Targeted Feature Mode Extraction Method With Adaptive Local Feature Enhancement in Rolling Bearing Fault Diagnosis

The increasing complexity of modern machinery has led to frequent occurrences of compound bearing faults with coupled characteristics, posing significant challenges for traditional diagnostic methods in industrial noise environments. This article proposes a targeted feature mode extraction method (TFME) with adaptive local feature enhancement. This method divides the frequency band through order statistics filters (OSFs) according to characteristics of the Fourier spectrum, so that the complete fault information can be obtained and the noise outside the passband of each initial mode is suppressed. Then a new estimated method of fault characteristic frequency, the envelope harmonic kurtosis spectrum (EHKS), is introduced. Based on this, an adaptive feature enhancement method (AFEM) is proposed so that the initial mode can be enhanced according to its own feature. By developing a structural parameter-guided targeted mode selection mechanism, this work effectively avoids the possibility of eliminating fault modes in conventional decomposition methods. Finally, the accuracy of extracting multiple fault features was verified through simulation and experiments.