Rolling Bearing Fault Diagnosis Based on SVDP-Based Kurtogram and Iterative Autocorrelation of Teager Energy Operator

The emergence of periodic impacts in the vibration signal is considered as an essential sign of rolling bearing faults. Therefore, how to distinguish the periodic impact component from the interference components (e.g., the harmonics and noise) in the raw vibration signal is critical for detecting bearing faults. The kurtogram technique plays an essential role in the automatic selection of sub-component signals containing fault information. However, two significant shortcomings reduce its ability to detect early weak transients: 1) the decomposition accuracy of the filters used in kurtogram, i.e., short-time Fourier transform (STFT) and binary filter banks, is deficiency and 2) the detection ability of kurtosis to cyclic impact is insufficient. A singular value decomposition packet (SVDP)-based kurtogram is proposed to improve the kurtogram technique. More specifically, a novel parameter-less signal decomposition algorithm, termed SVDP, is employed as the filter for sub-components extraction. The L-kurtosis indicator is then introduced to replace the kurtosis indicator to select the optimal sub-component from the SVDP processing results. Moreover, a fault signature highlighting technique named iterative autocorrelation of Teager energy operator (TEO-IAC) is presented, and the TEO-IAC spectrum is adopted to replace the Hilbert envelope spectrum to detect the fault characteristic frequencies of the optimal sub-component to determine the fault types of bearings. Finally, the presented fault diagnosis framework based on the SVDP-based kurtogram and TEO-IAC is compared with the original kurtogram, improved kurtogram and autogram in simulated and experimental signals analysis, which demonstrates its validity and superiority for extracting weak fault features of bearings.