Self-Matching Chirplet Extraction Transform: A Novel Tool for Dense Multicomponent Signals Analysis and Machinery Fault Diagnosis

This study introduces a novel time–frequency (TF) analysis methodology, designated as the self-matching chirplet extraction transform (SMCET). This innovative technique is specifically crafted for the analysis of nonstationary signals characterized by dense multicomponents, aimed at achieving an accurate TF representation (TFR). SMCET extends chirplet transform (CT) by employing parameterized phase kernels with higher order expansions, which accurately fits the instantaneous frequencies (IFs) of nonlinear variations, thereby achieving precise matching between chirprate and varying frequency. This technique effectively resolves the issue of TF resolution ambiguity, a challenge encountered by CT-improved algorithms, when processing dense components. The optimization strategy for self-matching parameter significantly enhances both precision of matching and computational efficiency by narrowing the traversal range of chirprate angle. By combining IF estimate with the matching extraction operator (MEO), a higher quality TFR can be obtained, even in the case of noise interference. The efficacy of SMCET is convincingly demonstrated via numerical and experimental analyses, focusing on vibration signals of bearing and planetary gearbox. The analytical outcomes reveal that SMCET exhibits superior capabilities in characterizing nonlinear multicomponent signals, including those with dense components. This underlines its significant potential for accurate fault diagnosis on rotary machinery.