Frequency autocorrelation mode decomposition for rotating machinery fault component extraction

Abstract Fault component extraction is crucial for rotating machinery fault diagnosis. However, the majority of existing signal processing methods designed for fault signal component extraction have limited precision and can only extract one fault component, ignoring the probably existing multiple fault components in different frequency bands. Therefore, it is necessary to design new signal processing methods to accurately extract multiple fault components. To meet this need, this article proposed a new signal decomposition method, namely frequency autocorrelation mode decomposition (FAMD). The new FAMD is based on an iterative frequency band searching model and a new dimensionless harmonic frequency autocorrelation index (HFAI) defined on a square envelope spectrum (SES). Thanks to the iterative searching model, the FAMD can automatically determine the number of fault components in different frequency bands and further extract and separate them. Besides, because the new HFAI is dimensionless, the FAMD can extract weak fault components. Two experimental cases demonstrated the effectiveness of the proposed FAMD, which is superior to the kurtogram, variational mode decomposition, and feature mode decomposition.